Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Analysis of interworking architectures for IP multimedia subsystem Munir, Arslan 2007

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Notice for Google Chrome users:
If you are having trouble viewing or searching the PDF with Google Chrome, please download it here instead.

Item Metadata

Download

Media
831-ubc_2007-0197a.pdf [ 8.11MB ]
Metadata
JSON: 831-1.0100992.json
JSON-LD: 831-1.0100992-ld.json
RDF/XML (Pretty): 831-1.0100992-rdf.xml
RDF/JSON: 831-1.0100992-rdf.json
Turtle: 831-1.0100992-turtle.txt
N-Triples: 831-1.0100992-rdf-ntriples.txt
Original Record: 831-1.0100992-source.json
Full Text
831-1.0100992-fulltext.txt
Citation
831-1.0100992.ris

Full Text

Analysis of Interworking Architectures for IP Multimedia Subsystem by Arslan Munir B.Sc , Electrical Engineering, University of Engineering and Technology, Lahore, 2004 A THESIS S U B M I T T E D IN P A R T I A L F U L F I L M E N T OF T H E R E Q U I R E M E N T S F O R T H E D E G R E E OF Master of Applied Science in The Faculty of Graduate Studies ( Electrical and Computer Engineering) The University of British Columbia April 2007 © Arslan Munir 2007 Abst rac t The future fourth generation wireless heterogeneous networks aim to integrate various wireless access technologies and to support the IMS (IP multimedia subsystem) ses-sions. In the first part of this thesis, we propose the Loosely Coupled Satellite-Cellular-W i M a x - W L A N ( L C S C W 2 ) and the Tight ly Coupled S a t e l l i t e - C e l l u l a r - W i M a x - W L A N ( T C S C W 2 ) interworking architectures. The L C S C W 2 and T C S C W 2 architectures use the loosely coupling and tightly coupling approach respectively and both of them integrate the satellite networks, 3rd generation (3G) wireless networks, worldwide interoperability for microwave access ( W i M a x ) , and wireless local area networks ( W L A N s ) . They can support IMS sessions and provide global coverage. The L C S C W 2 architecture facilitates independent deployment and traffic engineering of various access networks. The T C -S C W 2 aims to guarantee quality of service (QoS). We also propose an analytical model to determine the associate cost for the signaling and data traffic for inter-system com-munication in the L C S C W 2 and T C S C W 2 architectures. The cost analysis includes the transmission, processing, and queueing costs at various entities. Numerical results are presented for different arrival rates and session lengths. In the second part of this the-sis, the signaling flows for IMS registration, IMS session establishment and IMS session i i Abstract re-establishment procedure after undergoing a vertical handoff in a 4 G environment are analyzed. Signaling delays are calculated for IMS signaling taking into account trans-mission, processing and queueing delays at network entities. The proposed analysis of the IMS signaling flows is independent of a particular access network technology and interworking architecture and can be applied to any of the access network technology and 4 G interworking architecture. i i i Table of Contents Abstract ii Table of Contents iv List of Tables vii List of Figures viii List of Acronyms x Acknowledgements xvii 1 Introduction 1 1.1 Motivations 2 1.2 Contributions 2 1.3 Structure of the Thesis 3 2 Related Work 5 2.1 Literature Survey . . 5 2.2 IP Multimedia Subsystem (IMS) 12 iv Table of Contents 2.2.1 IMS architecture 13 The databases: the HSS and the S L F 14 The C S C F 14 2.3 4 G Interworking Architectures 16 2.3.1 Tight ly Coupled (TC) Interworking Architecture 16 2.3.2 Loosely Coupled (LC) Interworking Architecture 18 2.3.3 Hybr id Coupled (HC) Interworking Architecture 19 2.3.4 The T C D R A S Interworking Architecture 21 2.3.5 The L C D R A S Interworking Architecture 22 2.4 Summary 24 3 L C S C W 2 and T C S C W 2 : Architectures for IP Multimedia Subsystem 25 3.1 The L C S C W 2 Architecture 26 3.2 The T C S C W 2 Architecture 29 3.3 A n Analyt ica l Mode l for Cost Analysis 34 3.3.1 Available Paths for Communications 35 3.3.2 Transmission Cost 36 3.3.3 Processing Cost 38 3.3.4 Queueing Cost 40 3.4 Numerical Results 42 3.5 Summary 48 v \ Table of Contents 4 Analysis of SIP-based Signaling for IMS Sessions in 4G Networks 50 4.1 Delay Analysis of IMS Signaling Procedures 51 4.1.1 Transmission Delay 52 4.1.2 Processing Delay 60 4.1.3 Queueing Delay 62 4.1.4 Tota l Delay 64 IMS Registration Procedure Delay 65 IMS Session Setup Delay 65 IMS Session Re-establishment Delay 66 4.2 Numerical Results 67 4.3 Summary 72 5 Conclusions and Future Work 83 5.1 Conclusions 83 5.2 Future Work 84 Bibliography 85 vi List of Tables 4.1 Size of SIP messages involved in IMS signaling 55 4.2 Values of K for signaling messages in different channel rates 57 v i i List of Figures 2.1 3 G P P IMS architecture 13 2.2 Tight ly coupled interworking architecture 17 2.3 Loosely coupled interworking architecture 18 2.4 Hybr id coupled interworking architecture 20 2.5 The T C D R A S interworking architecture 22 2.6 The L C D R A S interworking architecture 23 3.1 The L C S C W 2 interworking architecture 26 3.2 The T C S C W 2 interworking architecture 30 3.3 Signaling and data communication paths in the L C S C W 2 architecture. . 35 3.4 Signaling and data communication paths in the T C S C W 2 architecture. . 36 3.5 The breakup of system signaling cost in the L C S C W 2 architecture. . . . 44 3.6 The breakup of system signaling cost in the T C S C W 2 architecture. . . . 45 3.7 Effect of varying arrival rate on system signaling cost 46 3.8 Effect of varying arrival rate on system data cost 47 3.9 Effect of varying session duration on system data cost 48 v i i i List of Figures 4.1 GPRS attach procedure 58 4.2 P D P context activation procedure 73 4.3 D H C P registration process 73 4.4 IMS registration process 74 4.5 IMS session setup procedure 75 4.6 IMS registration for fixed A, p 76 4.7 Session setup when SN is in 3G and C N is in W L A N for fixed A, p . . . . 76 4.8 Session setup when SN and C N are in 3G for fixed A, p 77 4.9 Session setup when SN and C N are in W L A N for fixed A, p 77 4.10 Session re-establishment when SN in 3G and C N in W L A N for fixed A, p 78 4.11 Session re-establishment when SN and C N are in 3G for fixed A, p . . . . 78 4.12 Session re-establishment when SN in W L A N and C N in 3G for fixed A, p 79 4.13 Session re-establishment when SN and C N are in W L A N for fixed A, p . 79 4.14 Effect of changing A on IMS registration delay for fixed p 80 4.15 Effect of changing A on IMS session setup delay for fixed p 80 4.16 Effect of changing A on IMS session re-establishment delay for fixed p . . 81 4.17 Effect of changing p on IMS registration delay for fixed A 81 4.18 Effect of changing p on IMS session setup delay for fixed A 82 4.19 Effect of changing p on IMS session re-establishment delay for fixed A . . 82 ix List of Acronyms 3 G 3 r d Generation 3 G P P 3 r d Generation Partnership Project A A A Authentication, Authorizat ion and Account A N Access Network A P Access Point A R G Access Router and Gateway A S Access Stratum B G C F Breakout Gateway Control Function B E R Bi t Error Rate B S C Base Station Controller C B R Constant B i t Rate C D M A Code Division Mul t ip le Access List of Acronyms C N Correspondent Node D A D Duplicate Address Detection D H C P Dynamic Host Configuration Protocol D N S Domain Name System DS Differentiated Services D S C P Differentiated Services Code Point E A P Extensible Authentication Protocol E I R Equipment Identification Register G G S N Gateway G P R S Support Node G P R S General Packet Radio Service G S M Global System for Mobile Communications H A R Q Hybr id Automat ic Request H C Hybr id Coupled HCSCW2 Hybr id Coupled S a t e l l i t e - C e l l u l a r - W i M a x - W L A N H L R Home Location Register H N A C Hotspot Network Area Controller List of Acronyms H S S H o m e Subscr iber Server I -CSCF In te r roga t ing-Ca l l Session C o n t r o l F u n c t i o n I E T F Internet Eng inee r ing Task Force I M S I P M u l t i m e d i a Subsys t em I P Internet P r o t o c o l I PSec I P Secur i t y I S D N Integrated Services D i g i t a l Ne twork I SL Inter-satel l i te L i n k I S U P I S D N User P a r t L C D R A S Loose l y C o u p l e d w i t h D i rec t R a d i o Access S y s t em L C S C W 2 Loose l y C o u p l e d S a t e l l i t e - C e l l u l a r - W i M a x - W L A N M A C M e d i u m Access C o n t r o l M G C F M e d i a G a t e w a y C o n t r o l F u n c t i o n M G W M e d i a Ga t eway M R F M e d i a Resource F u n c t i o n M R F C M e d i a Resource F u n c t i o n Con t ro l l e r List of Acronyms M R F P Media Resource Function Processor M T Mobile Terminal N A S Non Access Stratum N A T Network Address Translation O B P On-board Processing P C F Packet Control Function P - C S C F Proxy-Ca l l Session Control Function P D C H Packet Da ta Channels P D G Packet Da ta Gateway P D P Packet Da ta Protocol P D S N Packet Da ta Serving Node P D U Packet Da ta Uni t P I M - S M Protocol-Independent Multicast Sparse Mode P L M N Public Land Mobile Network PS Packet Switched P S T N Publ ic Switched Telephone Network List of Acronyms QoS Quali ty of Service R A Routing Area R B Radio Bearer R L C - A M Radio Link Control-Acknowledged Mode R L P Radio Link Protocol R N C Radio Network Controller R T O Retransmission Time Out R T T Round Trip Time R - U I M Removable-User Identity Module S 3 G I F Satellite-3G Interworking Function SigComp Signaling Compression S - C S C F Serving-Call Session Control Function S F E S Satellite Fixed Ear th Station S G S N Serving G P R S Support Node S G W Signaling Gateway S I M Subscriber Identity Module Lis t of Acronyms SIP Session Initiation Protocol S L A Service Level Agreement S L F Subscriber Location Function S M S Short Messaging Service S N Source Node SS7 Signaling System Number 7 SSB Satellite Spot Beam S - U M T S Sate l l i te -UMTS T C Tight ly Coupled T C D R A S Tight ly Coupled wi th Direct Radio Access System T C P Transmission Control Protocol T C S C W 2 Tight ly Coupled Sa t e l l i t e -Ce l l u l a r -WiMax-WLAN T E Terminal Equipment ToS Type of Service U D P User Datagram Protocol U E User Equipment xv List of Acronyms U M T S Un i ve r sa l M o b i l e Te l e commun i ca t i ons S y s t em U S I M Un i ve r sa l Subscr iber Ident i ty M o d u l e V B R Va r i ab l e B i t R a t e V o I P Vo ice over I P W G D C T W L A N to 3 G D i rec t Con t ro l l e r a n d Transce iver W A G Wi re less Access G a t e w a y W B S C W i M a x Base S t a t i o n Con t ro l l e r W i M a x W o r l d w i d e In te roperab i l i t y for M i c rowave Access W I F W L A N - 3 G In te rwork ing F u n c t i o n W L A N Wi re less L o c a l A r e a Ne two rk W M I F W i M a x - 3 G In te rwork ing F u n c t i o n W N C W i M a x Ne twork Con t ro l l e r x v i Acknowledgements I w o u l d l ike to express m y sincere g ra t i tude to m y gradua te superv isor D r . V i n c e n t W o n g for his gu idance a n d suppor t d u r i n g the course of m y graduate studies. I s incerely apprec ia te the cons iderab le amoun t of t ime a n d effort he invested i n he lp ing me w i t h m y research a n d thesis. I w o u l d l ike to t h a n k m y fel low col leagues, p a r t i c u l a r l y Syed Hussa in A l i , E n r i q u e Stevens-Navarro, a n d E h s a n B a y a k i , w h o have p rov ided he lp fu l suggestions a long the way. I also acknowledge D r . V i k r . am K r i s h n a m u r t h y whose gu idance and suggest ions he lped me to improve the overa l l h o r i z o n of m y knowledge. T h i s work is suppo r t ed by B e l l C a n a d a a n d N a t u r a l Sciences a n d Eng inee r i ng R e -search C o u n c i l of C a n a d a under pro jec t grant 328202-05. x v i i Chapter 1 Introduction The 4th generation (4G) wireless heterogeneous networks are envisioned as the integra-tion of various wireless access technologies such as wireless local area network ( W L A N ) , 3rd generation (3G) technologies including universal mobile telecommunications system ( U M T S ) and code division multiple access ( C D M A ) based C D M A 2 0 0 0 system, worldwide interoperability for microwave access ( W i M a x ) , and satellite networks. The aim behind this integration is to provide the users wi th global coverage and to provide the users wi th the capability to switch between different available access networks (ANs) . The success of the 4 G wireless heterogeneous networks depends on the successful integration of the currently available A N s . The IP multimedia subsystem (IMS) is standardized by the 3rd generation partner-ship project ( 3GPP) and 3 G P P 2 as a new core network domain [1]. The IMS enables the provision of internet protocol (IP) based multimedia applications to mobile users, guarantees quality of service (QoS) across different A N technologies and permits service providers to charge according to different policies. In addition, the IMS enables third-party vendors to develop new applications for operators and users. A growing number of telecommunication vendors are beginning to release devices and services based on the 1 Chapter 1. Introduction IMS [2]. 1.1 Motivations Although, some 4G interworking architectures have been proposed before which integrate 3G and W L A N or 3G and satellite networks or 3G and W i M a x individually, to the best of our knowledge, a 4G interworking architecture that integrates satellite, W i M a x , and W L A N with the 3G network has not been proposed so far. Addit ionally, the IMS infras-tructure has not been incorporated in the previously proposed interworking architectures. The research related to performance evaluation of the interworking architectures so far either analyzes the performance or cost for signaling traffic or data traffic. Li t t le or no work has been done before that evaluates the system performance for both signaling and data. Also a comprehensive cost analysis taking into account the transmission, processing and queueing costs of traffic in the interworking architectures has not been done before. Addit ionally, a comprehensive delay analysis of SIP-based IMS registration, IMS session setup, and IMS session re-establishment process has not been performed before. 1.2 Contributions The main contributions of this thesis are given below: • We propose two novel 4G interworking architectures namely the Loosely Coupled S a t e l l i t e - C e l l u l a r - W i M a x - W L A N (LCSCW2) architecture and the Tight ly Coupled 2 Chapter 1. Introduction S a t e l l i t e - C e l l u l a r - W i M a x - W L A N ( T C S C W 2 ) architecture based on loosely coupled and tightly coupled paradigms respectively. The IMS has been given special con-sideration in the proposed interworking architectures. The proposed architectures integrate 3G, W i M a x , W L A N , and satellite networks. • We also propose a cost analysis model for the signaling and data traffic for inter-system communication in the proposed interworking architectures. The cost anal-ysis includes the transmission, processing and queueing costs at various network entities. Our analysis takes into account both signaling and data traffic and de-scribes the effect of changing traffic characteristics such as arrival rates and IMS session duration on system cost. • We analyze the signaling flows involved in the IMS session establishment and reg-istration considering signaling compression (SigComp) for reducing the delay. We also analyze different scenarios for IMS session re-establishment after vertical hand-off in a 4 G environment. 1.3 Structure of the Thesis This thesis is structured as follows. In Chapter 2, related work and the 4 G interwork-ing architectures previously proposed are described. Chapter 3 describes our proposed L C S C W 2 and T C S C W 2 interworking architectures. Chapter 4 presents the analysis of SIP-based signaling for IMS sessions in the 4 G networks. Chapter 5 summarizes the main 3 Chapter 1. Introduction contributions of the thesis, and describes future trends in the area. 4 Chapter 2 Related Work 2.1 Literature Survey In this section, we begin by describing the previously done work in the area. Two W L A N -3 G interworking architectures have been described in [3]. One architecture can provide 3G-based access control and charging whereas the other can provide 3G-based access control and charging as well as access to 3 G packet switched (PS) based services. A d d i -tionally, authentication, authorization, and accounting ( A A A ) signaling required for the two discussed architectures is given. A n integrated U M T S IMS architecture is presented in [4] and the signaling flows for IP multimedia session control are described. The paper emphasizes the point that when multiple media types are involved in a session e.g. video and audio, synchronization is essential for simultaneous presentation of media types to the user. The IMS media gateway ( M G W ) is responsible for media signals translation between different formats when IMS session is established between two different A N s . The IMS media gateway is controlled by the media gateway control function ( M G C F ) which provides application-level signaling translation, for eg. between session init iation protocol (SIP) and integrated services digital network (ISDN) user part (ISUP) signal-5 Chapter 2. Related Work ing which is the case when one A N is public switched telephone network ( P S T N ) and the other is U M T S . The signaling gateway (SGW) performs the transport-level signaling translation between IP-based and Signaling System Number 7 (SS7) based transport. The performance evaluation of three U M T S - W L A N interworking strategies namely mobile IP approach, gateway approach, and emulator approach has been done in [5] and the signaling flows for U M T S to W L A N and W L A N to U M T S handover are given for the three strategies. The mobile IP approach introduces mobile IP to the two networks. M o -bile IP mechanisms are implemented in the mobile nodes and on the U M T S and W L A N network devices. This approach provides IP mobility for the roaming between U M T S and W L A N . The gateway approach introduces a new logical node that connects the two wireless networks. The node exchanges necessary information between the networks, con-verts signals, and forwards the packets for the roaming users. Through this approach, the handoff delay and packet loss can be reduced. The emulator approach uses W L A N as an access stratum (AS) in the U M T S network. It replaces the U M T S A S by the W L A N layer one and layer two. A W L A N access point (AP) can be viewed as serving general packet radio service ( G P R S ) support node (SGSN) . The advantage of this approach is that mobile IP is no longer required. A l l packet routing and forwarding are processed by U M T S core network. The packet loss and delay can be significantly reduced by this approach. The voice over IP (VoIP) performance in 3 G - W L A N interworking system wi th IP security (IPSec) tunnel between packet data gateway ( P D G ) and user equipment (UE) 6 Chapter 2. Related Work is evaluated in [6]. The performance metrics considered are packet inter-arrival time, data transmission rate, end-to-end.delay, and packet loss. The user mobility in and out of W L A N is considered and its effect on end-to-end delay is discussed. It is observed that as the number of VoIP connections at the A P increases, the delay increases to an unacceptable level and the performance of al l VoIP connections in that particular A P degrades drastically because the A P is unable to pol l all the clients in the point coor-dination function mode. A 3 G P P - W L A N interworking architecture in which subscriber identity module (SIM)-based authentication is used is described in [7]. Da ta routing is described for a user in W L A N accessing IP-based services in some external network. Charging infrastructure for 3 G P P - W L A N interworking architecture is described. Radius, Diameter, and extensible authentication protocol ( E A P ) are discussed for authentication purposes in the interworked architecture. A n interworking architecture is proposed in [8] in which the G P R S network is available all the time forming a primary network and W L A N s are used as a complement when they are available. The control part never leaves the U M T S and hence there is no need for control procedures in the W L A N i.e. paging and assignment of cell or routing area identifiers. W i t h the proposed architecture, IP sessions can be maintained in the hotspot network dark areas and short messaging service (SMS) can be accessed via W L A N . In the proposed architecture, no core network interfaces are exposed to the exterior of the core and hence ensures security. W i t h the proposed architecture, no information is ever lost in handovers because the hotspot network area controller ( H N A C ) wi l l transmit the 7 Chapter 2. Related Work packet again through the other interface. The loose coupling and tight coupling interworking architectures are discussed in [9]. It is discussed that the U E is made up of two disjoint entities, i.e. terminal equipment (TE) and mobile terminal ( M T ) according to 3 G P P specifications. Different scenarios in U M T S - W L A N interworking are described. The one is that in which W L A N gateway is responsible for both the non access stratum (NAS) and A S signaling. The other is that in which the T E handles the N A S signaling. Another one is that in which 802.11 nodes can be configured in ad-hoc mode and communicate wi th one another directly and the gateway is required only for accessing U M T S services. A n architecture for integrating C D M A 2 0 0 0 and 802.11 W L A N is presented in [10]. The architecture is tightly coupled since it proposes the re-use of packet data serving node ( P D S N ) of C D M A 2 0 0 0 network for W L A N traffic forwarding as well. The signaling flows when the U E is powered up in a W L A N , and C D M A 2 0 0 0 to W L A N handover procedure are given. The possibility of integration of satellite with terrestrial systems has been discussed in [11]. It is pointed out that satellite coverage can work alone in air and sea but the success of the satellite systems in land areas lies in their integration wi th terrestrial systems. On-board Processing ( O B P ) enables satellite systems to interconnect satellite spot beams (SSBs) and allows variable bandwidth channels. O B P allows dynamic routing between various spot beams and provides support for real-time applications such as VoIP and multiparty conference services. Satellites can be interconnected in an orbit v ia inter-satellite link (ISL). One SSB covers many 3G cells and is suitable for high velocity users. 8 Chapter 2. Related Work In the satellite-3G integrated system, the number of handoffs for a fast moving user wi l l be minimized and hence the probability of handoff call dropping is reduced. The integrated satellite-3G system can be used to offload congestion in the 3 G network by handing off the users from 3 G to satellite system. The footprints of SSBs cover many cellular cells and provides a pool of channels to be shared by these cells. The idea of using satellite capacity to mitigate congestion in areas served by terres-tr ia l network has been explained in [12]. The paper evaluates the performance of the satellite-terrestrial integrated system by considering a one-dimensional analytical model of a cellular system overlaid with satellite footprints. The paper also simulates planar cellular network with satellite spot beam coverage support. It is shown that the inte-grated system can improve Erlang-B blocking performance. The SIP based session setup signaling for Sate l l i te -UMTS ( S - U M T S ) is discussed in [13] based on the current radio link control acknowledged mode ( R L C - A M ) . The paper proposes two schemes to reduce the inefficiency of the current R L C - A M mechanism for session setup over S - U M T S . The first scheme can recover the missing last radio segments in a single round trip time and the second scheme can reduce the redundant transmissions which occur due to multiple feedback triggers. A n architecture for multiparty conferencing over satellites is described in [14]. A SIP-based conference signaling and an extension to protocol independent multicast-sparse mode ( P I M - S M ) that supports QoS in DiffServ networks is proposed. A satellite emulator is used to obtain the results of user perceived QoS, signaling delay, and jitter. It is observed that the delay at 75 % background traffic is longer than that 9 Chapter 2. Related Work at 25 % background traffic before the activation of QoS mechanisms but it becomes the same after the activation of QoS mechanisms. A S - U M T S architecture is presented in [15] and possible signaling flows for registration, call handling and handover are given. It is discussed that for U E initiated calls, the satellite earth station allocates the resources which are one or more packet data channels ( P D C H ) that are shared by the users covered by a SSB. The signaling flows for conference creation over S - U M T S are described in [16] and a simulation model is presented for S - U M T S . The results show that the conference creation delay increases wi th the increasing block error rate and the resource reservation delay is the main contributing factor in the total delay. Also, use of packet data units (PDUs) of large size reduces the block error probability and hence the delay. A n overview of W i M a x / I E E E 802.16 is provided in [17]. The paper mentions the achievable throughput of W i M a x and suggest some enhancements to the I E E E 802.16 standard that have the potential of achieving higher data rates. A n architecture for U M T S - W i M a x interworking is proposed in [18] and signaling flows of handover from W i M a x to U M T S access network and vice versa are given. The paper describes the difference between U M T S - W L A N interworking and U M T S - W i M a x interworking. The U M T S and W L A N are fully overlapped because when a U E is connected to W L A N , it can maintain simultaneous connection to the U M T S network. The U M T S and W i M a x are partially overlapped because W i M a x coverage area is in order of U M T S coverage area and simultaneous connection to the two A N s is not possible at al l the times. In [19], an architecture is proposed for integration of W i M a x and U M T S based on loosely-coupled 10 Chapter 2. Related Work approach. A mapping is provided between the QoS classes in U M T S and W i M a x . The paper examines the throughput for constant bit rate ( C B R ) voice application and variable bit rate ( V B R ) video application v ia simulations. The signaling efficiency for call setup in I M S infrastructures using C D M A 2 0 0 0 is analyzed in [20]. Signaling flows involved in call establishment procedure are shown. It is assumed that both the S N and the C N are in C D M A 2 0 0 0 system. The effect of early termination due to the hybrid automatic request ( H A R Q ) algorithm is considered in the forward link. Four call setup scenarios are considered. The first one is trans-mission control protocol ( T C P ) based; the second assumes additional encoding of SIP messages apart from SigComp so that every SIP packet fits in 1 single radio frame; the third assumes that proxy-call session control function ( P - C S C F ) of caller can directly communicate with the callee P - C S C F ; the fourth one uses a variation of user datagram protocol ( U D P ) in which U D P wi th retransmission time out (RTO) is used, moreover R T O value is set equal to round trip time ( R T T ) value. The SIP session setup delay for VoIP service in 3G wireless networks is studied in [21, 22]. A n adaptive retransmission timer is considered for retransmission of lost packets at the application layer. The effect of T C P , U D P , and radio link protocol ( R L P ) is considered on SIP session setup for VoIP. Performance of SIP-based vertical handoff is analyzed in [23]. Signaling flows are given for G P R S attach procedure, packet data protocol ( P D P ) context activation procedure, SIP-based mid-call terminal mobility, and dynamic host configuration protocol ( D H C P ) registration procedure. Analy t ica l expressions for delay of SIP-based handoff to U M T S 11 Chapter 2. Related Work network from another U M T S network or a W L A N as well as SIP-based handoff to W L A N network from another W L A N or a U M T S network are given. SIP-based mobility in IPv6 is described in [24]. The paper closely examines the delay incurred when a U E moves to a new link and performs the duplicate address detection ( D A D ) and router selection. It is shown that intelligent modifications to IPv6 Linux kernel implementation achieve a faster handoff in SIP-based terminal mobili ty as compared to unaltered Linux kernel. 2.2 IP Multimedia Subsystem (IMS) In 3 G P P / 3 G P P 2 , the IMS is defined as an architectural infrastructure developed wi th the intent of delivering IP multimedia services to the end-users. This framework requires to meet the following requirements [25], [1]: • Support for establishing IP multimedia sessions. • Support for mechanism to arbitrate QoS. • Support for interworking between Internet and circuit-switched networks. • Support for roaming. • Support for powerful control imposed by the operator with respect to the services provided to the end-users. • Support for expeditious service development without requiring standardization. • Support for access from other networks (known as access independence in IMS) . 12 Chapter 2. Related Work MRFC r'AccessN UE P-CSCF UE: User Equipment HSS: Home Subscriber Server l-CSCF SLF: Subscriber Location Function MGW: Media Gateway AS: Application Server MRFC: Media Resource Function Controller SGW: Signalling Gateway MRFP: Media Resource Function Processor M r T u u MGCF: Media Gateway Control Function BGCF: Border Gateway Control Function M t , w P/S/I-CSCF: Proxy/Serving/Interrogating - Call Session Control Function Figure 2.1: 3 G P P IMS architecture [25]. 2.2.1 IMS architecture Here, we present an overview of the IMS architecture which is benefited from [25], [1] as shown in Figure 2.1. It is worth mentioning that 3 G P P does not standardize nodes in IMS, but functions. This implies that the IMS architecture is a collection of func-tions linked by standardized interfaces. The nodes included in the IMS are one or more user databases, called home subscriber servers (HSSs) and subscriber location functions (SLFs); one or more SIP servers, jointly referred to as C S C F s (Cal l Session Control Func-tions); one or more application servers; one or more M R F s (Media Resource Functions), each one is further characterized into M R F C (Media Resource Function Controllers) and 13 Chapter 2. Related Work M R F P (Media Resource Function Processor); one or more B G C F s (Breakout Gateway Control Functions); one or more P S T N gateways, each one is characterized into a S G W (Signaling Gateway), a M G C F (Media Gateway Control Function), and a M G W (Media Gateway). We briefly explain the functionalities of some of the entities in the core IMS network. The databases: the HSS and the SLF The HSS is the central database for user-related information. The HSS in essence is an evolution of the home location register ( H L R ) , which is a G S M node. The HSS ac-commodates al l the user-related subscription data needed to handle multimedia sessions. These data comprise mainly the location information, security information (including both authentication and authorization information), user profile information (including the services that the user is subscribed to), and the S - C S C F (Serving-CSCF) assigned to the user. Networks wi th a single HSS do not require an S L F . However, the networks with more than one HSS do require an S L F . The S L F is a database that is responsible for mapping users' addresses to HSSs. A node that queries the S L F , with the user's address as the input, acquires the HSS that has al l the information related to that user as the output. The C S C F The C S C F is a SIP server and processes SIP signaling in the IMS. The C S C F s can be characterized into three categories depending upon their functionality: 14 Chapter 2. Related Work • P - C S C F (P roxy-CSCF) • I - C S C F (Interrogating-CSCF) • S - C S C F (Serving-CSCF) The P - C S C F is the first point of contact (in the signaling plane) between the I M S terminal and the IMS network. From the SIP perspective, the P - C S C F is serving as an outbound/inbound SIP proxy server as al l the requests generated by the IMS terminal or destined to the I M S terminal traverse the P - C S C F . The P - C S C F is assigned to the I M S terminal during IMS registration and does not change for the duration of the registration. The I - C S C F is a SIP proxy located at the edge of an administrative domain. The address of the I - C S C F is listed in the D N S (Domain Name System) records of the domain. When a SIP server follows standard SIP procedures to determine the next SIP hop for a specific message, the SIP server acquires the address of an I - C S C F of the destination domain. The I - C S C F is responsible for choosing the appropriate S - C S C F based on load or other capability. Apar t from the SIP proxy server functionality, the I - C S C F has an interface to the S L F and the HSS. The I - C S C F fetches user location information and directs the SIP requests to the suitable S - C S C F . The S - C S C F is basically a SIP server and it executes session control as well. S - C S C F preserves a binding between the user location (e.g., the IP address of the terminal the user is logged on) and the user's SIP address of record (also known as public user identity) and therefore acts as a SIP registrar. Like the I - C S C F , the S - C S C F is also connected to 15 Chapter 2. Related Work the HSS in order to download the authentication vectors of the user and the user profile. It also informs the HSS that this is the S - C S C F assigned to the user for the duration of the registration. 2.3 4G Interworking Architectures 2.3.1 Tightly Coupled (TC) Interworking Architecture The tightly coupled (TC) W L A N - U M T S interworking architecture is shown in Figure 2.2. We use dotted lines between the two entities in our architecture to show that the two entities only share signaling messages with each other and data packets never flow through these entities. The dotted circles around A P s of W L A N and base station controllers (BSCs) of 3G represent their respective coverage areas. A n overlay architecture which is the main essence of 4 G networks is considered in which W L A N users are also covered by B S C of 3G. The discussion has been benefited from [9, 26, 27]. The main concept behind the tightly-coupled approach is to give W L A N appearance of another U M T S access network from the perspective of U M T S core network. In other words, the W L A N is considered like another G P R S routing area ( R A ) in the system. The W L A N - 3 G interworking function (WIF) , which is connected to the S G S N of the U M T S core network, is responsible for hiding the details of the W L A N from the U M T S core network and implementation of the U M T S protocols for mobili ty management, authentication etc. essential for the U M T S radio access network. For seamless operation in the interworked 16 Chapter 2. Related Work P-CSCF ARG: Access Router & Gateway PDN: Packet Data Network GGSN: Gateway GPRS Support Node SGSN: Serving GPRS Support Node BSC: Base Station Controller PDG: Packet Data Gateway IMS: IP Multimedia Subsystem RNC: Radio Network Controller WAG: Wireless Access Gateway PDSN: Packet Data Serving Node HSS: Home Subscriber Server P& Packet Switched UE: User Equipment AP: Access Point PCF: Packet Control Function WIF: WLAN-3G Interworking Function 3GPP: 3 f d Generation Partnership Project AAA: Authentication, Authorization and Accounting Figure 2.2: Tight ly coupled interworking architecture. architecture, U E s are required to implement the U M T S protocol stack on the top of their standard I E E E 802.11 W L A N network cards. Among the disadvantages of tightly-coupled approach is the exposure of the U M T S core network interfaces directly to the W L A N network which invites security challenges. Extensive efforts are required for the implementation of W I F especially for the W L A N not owned by the U M T S operators. 17 Chapter 2. Related Work P-CSCF ARG: Access Router & Gateway PDN: Packet Data Network GGSN: Gateway GPRS Support Node SGSN: Serving GPRS Support Node BSC: Base Station Controller IMS: IP Multimedia Subsystem RNC: Radio Network Controller WAG: Wireless Access Gateway PDSN: Packet Data Serving Node HSS: Home Subscriber Server UEf User Equipment AP: Access Point PCF: Packet Control Function WLAN3GIF: WLAN-3G Interworking Function 3GPP: 3rd Generation Partnership Project AAA: Authentication, Authorization and Accounting Figure 2.3: Loosely coupled interworking architecture. 2.3.2 Loosely Coupled (LC) Interworking Architecture The loosely coupled (LC) W L A N - U M T S interworking architecture is shown in Figure 2.3. The discussion has been benefited from [26, 27]. In the loosely-coupled inter-working ar-chitecture, the W L A N connects to the external packet data network (actually connects to P - C S C F in case of IMS) and does not have any direct link to 3 G network elements such as SGSNs or gateway G P R S support nodes (GGSNs) . Loosely-coupled architec-ture has distinct data paths for W L A N and U M T S traffic. The inter-operability wi th 18 Chapter 2. Related Work 3G requires the support of Mobile-IP functionalities or SIP protocol to handle mobility across networks and A A A services in the wireless access gateway ( W A G ) of W L A N . This support is necessary to interwork with the 3G's home network A A A servers. One of the major advantages of the loosely-coupled architecture is that it permits independent de-ployment and traffic engineering of W L A N and 3G networks. The loose coupling utilizes standard Internet engineering task force ( I E T F ) based protocols for A A A and mobili ty and therefore, it does not necessitate the introduction of cellular technology into the W L A N network. The W L A N uses the E A P for authentication of the U E by supplying the subscriber identity, SIM-based authentication data, and encrypted session keys. In case when W L A N is not owned by the 3G operator and SIM-based authentication is not feasible in the W L A N system, standard user name and password procedures may be de-ployed. The service continuity during handover to other access networks is not supported efficiently in loose-coupling and causes significant handover latency and packet loss. 2.3.3 Hybrid Coupled (HC) Interworking Architecture The hybrid coupled (HC) W L A N - U M T S interworking architecture is shown in Figure 2.4. The basic H C interworking architecture is proposed in [28]. The H C architecture distin-guishes the data paths according to the type of traffic and is capable of accommodating traffic from W L A N efficiently wi th guaranteed QoS and seamless mobility. The tightly-coupled network architecture is chosen for real-time traffic and the loosely-coupled archi-tecture is selected for non-real time and bulky traffic. In the H C architecture, the access 1 9 Chapter 2. Related Work Data + Signaling Signaling only IMS Backbone Network S-CSCF S-CSCF P-CSCF GGSN/PDSN 3GPP AAAJ Server J j l HSS P-CSCF WIF ARG: Access Router & Gateway PDN: Packet Data Network GGSN: Gateway GPRS Support Node SGSN: Serving GPRS Support Node BSC: Base Station Controller PDG: Packet Data Gateway IMS: IP Multimedia Subsystem RNC: Radio Network Controller WAG: Wireless Access Gateway PDSN: Packet Data Serving Node . A P „ U E •,WAG% PDG l-CSCF [3GPP PS Services] ^including access to ] Internet Intranet/Internet \ UE PS: Packet^witched UE: User Equipment HSS: Home Subscriber Server AP: Access Point PCF: Packet Control Function WIF: WLAN-3G Interworking Function 3GPP: 3rd Generation Partnership Project AAA: Authentication, Authorization and Accounting Figure 2.4: Hybr id coupled interworking architecture. router and gateway ( A R G ) of W L A N is responsible for forwarding packets to/from an A P from/to the S G S N ; supporting interfaces for signaling and data traffic to/from S G S N ; management of radio resources in the W L A N and mapping them onto radio resources in the 3 G network; setting data path to S G S N or external packet data network based on the type of traffic where the service types are differentiated by type of service (ToS) or differ-entiated services (DS) code point ( D S C P ) field of the IP header. The Mobile-IP or SIP protocol may be used for seamless mobility in the H C architecture. The vertical handoff procedures in the H C architecture are similar to the procedures in T C architecture. In 20 Chapter 2. Related Work the H C architecture, the U E belongs to the U M T S network only for a user using U M T S access network. However, for a user accessing IMS services through W L A N belongs to both the U M T S and W L A N for efficient handover purposes. Hence, when a user moves from the W L A N to the U M T S , the registration to the U M T S network is not required because it is already registered. However when moving to the W L A N from U M T S , the registration procedure with the W L A N is required. 2.3.4 The T C D R A S Interworking Architecture The tightly coupled with direct radio access system ( T C D R A S ) W L A N - U M T S interwork-ing architecture is shown in Figure 2.5. The basic T C D R A S architecture is proposed in [29]. The T C D R A S is based on the tight-coupling architecture but it also creates an ad-ditional wireless link between the base station of a U M T S cell and W L A N located within the U M T S cell through W L A N to 3 G direct controller and transceiver ( W G D C T ) . The T C D R A S has a W i M a x transceiver collocated wi th W G D C T for direct wireless link be-tween W L A N and the U M T S base station. The signaling in T C D R A S is sti l l routed through the original T C path for network security purposes. However, it provides the opportunity for dynamic distribution of data traffic between the original T C path and the path available due to the added wireless link. The W G D C T uses I E E E 802.16 air interface and a W i M a x transceiver for direct wireless link establishment between W L A N and the U M T S base station. However, the cost of deployment of W i M a x transceivers in W L A N for directly forwarding the packets to the B S C may not be justified versus the 21 Chapter 2. Related Work S-CSCF PDN: Packet Data Network GGSN: Gateway GPRS Support Node -SGSN: Serving GPRS Support Node BSC: Base Station Controller PDG: Packet Data Gateway IMS: IP Multimedia Subsystem RNC: Radio Network Controller WAG: Wireless Access Gateway PDSN: Packet Data Serving Node PCF: Packet Control Function HSS: Home Subscriber Server PS: Packet Switched UE: User Equipment AP: Access Point ARG: Access Router & Gateway WIF: WLAN-3G Interworking Function 3GPP: Third Generation Partnership Project WGDCT: WLAN to 36 Direct Controller & Transceiver AAA: Authentication, Authorization and Accounting Figure 2.5: The T C D R A S interworking architecture. benefit they provide of reducing packet losses. 2.3.5 The L C D R A S Interworking Architecture The loosely coupled with direct radio access system ( L C D R A S ) W L A N - U M T S interwork-ing architecture is shown in Figure 2.6. The basic L C D R A S architecture is proposed in [30]. The L C D R A S is based on the loose-coupling architecture but it also creates an ad-ditional wireless link between the base station of a U M T S cell and W L A N located within the U M T S cell through W G D C T . The L C D R A S has a W i M a x transceiver co-located 22 Chapter 2. Related Work S - C S C F P-CSCF PDN: Packet Data Network '--.^ " " GGSN: Gateway GPRS Support Node ' — _ " " " " SGSN: Serving GPRS Support Node BSC: Base Station Controller IMS: IP Multimedia Subsystem HSS: Home Subscriber Server ARG: Access Router & Gateway RNC: Radio Network Controller UE: User Equipment 3GPP: Third Generation Partnership Project WAG: Wireless Access Gateway AP: Access Point WGDCT: WLAN to 3G Direct Controller & Transceiver PDSN: Packet Data Serving Node PCF: Packet Control Function AAA: Authentication, Authorization and Accounting Figure 2.6: The L C D R A S interworking architecture. with W G D C T for direct wireless link between W L A N and the U M T S base station. The I E E E 802.16 standard air interface is utilized in L C D R A S to set up direct wireless link between the base station in U M T S cellular networks and local W L A N . The L C D R A S is capable of dynamically distributing signaling and data traffic to reduce signaling cost and handoff latency via W G D C T . Here again, the cost of deployment of W i M a x transceivers in W L A N for directly forwarding the packets to the B S C may not be justified versus the benefit they provide of reducing packet losses. 23 Chapter 2. Related Work 2.4 Summary In this chapter, we presented the previously done work in the literature related to 4 G in-terworking architectures and IMS signaling analysis. Some of the proposed architectures in the literature are discussed in detail and modified to incorporate the IMS infrastruc-ture. 24 Chapter 3 LCSCW2 and TCSCW2: Architectures for IP Multimedia Subsystem In this chapter, we propose the L C S C W 2 and T C S C W 2 interworking architectures. The L C S C W 2 and T C S C W 2 architectures integrate the satellite networks, 3 G wireless net-works, W i M a x , and W L A N s , and are based on the loosely coupling and tightly coupling paradigm, respectively. They can support IMS sessions and provide global coverage. The L C S C W 2 architecture facilitates independent deployment and traffic engineering of var-ious access networks. We also propose an analytical model to determine the associate cost for the signaling and data traffic for inter-system communication in the L C S C W 2 architecture. The analytical model can be easily extended to determine the associated cost for the signaling and data traffic for inter-system communication in the T C S C W 2 architecture as well. The cost analysis includes the transmission, processing, and queue-ing costs at various entities. Numerical results are presented for different arrival rates and session lengths. 25 Chapter 3. LCSCW2 and TCSCW2: Architectures for IP Multimedia Subsystem PDN: Packet Data Network GGSN: Gateway GPRS Support Node SGSN: Serving GPRS Support Node BSC: Base Station Controller IMS: IP Multimedia Subsystem RNC: Radio Network Controller WAG: Wireless Access Gateway PDSN: Packet Data Serving Node ARG: Access Router & Gateway HSS: Home Subscriber Server WBSC: WiMax Base Station Controller WNC: WiMax Network Controller SFES: Satellite Fixed Earth Station ISL: lriter*Sate|lite Link 3GPP: Third Generation Partnership Project SSB: Satellite Spot Beam UE: User Equipment SFES AP: Access Point PCF: Packet Control Function AAA: Authentication, Authorization and Accounting P/S/I-CSCF: Proxy/Serving/lnterrogating-Call Session Control Function Figure 3 . 1 : The L C S C W 2 interworking architecture. 3.1 The LCSCW2 Architecture Our proposed L C S C W 2 interworking architecture is depicted in Figure 3 . 1 . This inter-working architecture integrates satellite networks, 3 G wireless cellular networks, W i M a x , and W L A N s based on the loosely coupled approach. The areas covered by the SSBs. 3 G base stations, W i M a x base stations, and W L A N A P s are shown by dotted lines in Figure 3 . 1 . Our proposed architecture is compatible with the IMS. Different A N s (e.g., 3 G networks, satellite networks, W i M a x , and W L A N s ) can be owned by different service providers (or the same operator). The W A G s of W L A N , W i M a x , satellite and 3 G net-26 Chapter 3. LCSCW2 and TCSCW2: Architectures for IP Multimedia Subsystem works are connected to different P - C S C F servers in IMS v ia the Internet. In general, each A N has its own separate W A G . In addition, there are separate serving-call session control function (S -CSCF) and interrogating-call session control function ( I -CSCF) servers for the two networks. For the establishment of an IMS session between two access networks, the two service providers should have a service level agreement (SLA) wi th each other. The I M S networks which are owned by different operators are connected together through an IMS backbone network. The W A G and P D S N are connected to the same P - C S C F server if W L A N , W i M a x , satellite and 3G network are owned by the same operator. The mechanisms involved in the interworking architecture along wi th the function-alities of various entities are explained below with reference to the 3 G P P specification [31]. Access to a locally connected IP network from a W L A N directly is called " W L A N direct IP access", which is provided by the loosely coupled architecture. The W A G is a gateway v ia which the data to/from the satellite A N , W i M a x A N , or W L A N A N can be routed to/from an external IP network. In the L C S C W 2 architecture, the satellite A N comprises of satellites and satellite fixed earth station (SFES) . Satellites convey data and signaling messages between U E and S F E S . The S F E S performs power control, link control, radio bearer control and paging functions. The S F E S is connected to W A G for accessing 3 G P P packet switched (PS) and I M S services. The S F E S determines whether or not a given U E can receive the requested services based on the user subscription. The S F E S also determines the U E ' s location v ia looking at the U E ' s entry in the database and its associated SSB ID. The U E ' s database entry is created when the U E first regis-27 Chapter 3. LCSCW2 and TCSCW2: Architectures for IP Multimedia Subsystem ters wi th the satellite network and is updated when the U E moves from coverage area of one SSB to another. The S F E S sends a page request to the U E including its identifier. During a pre-defined time, if the S F E S gets the response from the U E , it then instructs the U E to go through radio bearer (RB) establishment procedure. If the U E is unable to respond to the page message within the pre-defined time, the S F E S concludes that the U E is unreachable [15]. The W i M a x A N consists of W i M a x base stations, which are controlled by the W i M a x base station controller ( W B S C ) . Several W B S C s are controlled by one W i M a x network controller ( W N C ) . The W N C is connected to W A G to provide W i M a x users wi th 3 G P P P S and IMS services. The L C S C W 2 interworking architecture integrates satellite networks, 3 G wireless networks, W i M a x , and W L A N s based on the loose coupling approach since these A N s connect to the Internet or Intranet v ia W A G . Then, through the Internet or Intranet, the U E can access C S C F servers of the IMS network. In the L C S C W 2 architecture, the satellite network, W i M a x and W L A N do not have any direct link to 3 G network elements such as SGSNs or G G S N s . The L C S C W 2 ar-chitecture has distinct signaling and data paths for different A N s . The inter-operability with 3 G requires the support of mobile-IP functionalities and SIP to handle mobility across networks, and A A A services in the W A G of the A N . This support is necessary to interwork wi th the 3G's home network A A A servers. The authentication in the A N s is provided through the 3 G P P system [31]. The main advantage of the L C S C W 2 architec-ture is that it allows independent deployment and traffic engineering of satellite networks, 28 Chapter 3. LCSCW2 and TCSCW2: Architectures for IP Multimedia Subsystem W i M a x , W L A N s , and 3 G networks. In addition, this architecture utilizes standard I E T F based protocols for A A A and mobility in the W i M a x , W L A N s , and satellite networks. Our discussion and the cost analysis is equally valid whether the 3G technology be-i n g used is U M T S or C D M A 2 0 0 0 because there are corresponding network entities in C D M A 2 0 0 0 for the entities in U M T S . The P D S N in C D M A 2 0 0 0 performs the same functions as G G S N in U M T S . The packet control function ( P C F ) in C D M A 2 0 0 0 per-forms the same function as S G S N in U M T S . P C F connects to the P D S N in C D M A 2 0 0 0 as S G S N connects to G G S N in U M T S . 3.2 The T C S C W 2 Architecture Our proposed T C S C W 2 interworking architecture is depicted in Figure 3.2. This inter-working architecture integrates satellite networks, 3G wireless cellular networks, W i M a x , and W L A N s based on the loosely coupled approach. The areas covered by the SSBs, 3 G base stations, W i M a x base stations, and W L A N A P s are shown by dotted lines in Figure 3.2. Our proposed architecture is compatible with the IMS. Different A N s (e.g., 3G networks, satellite networks, W i M a x , and W L A N s ) can be owned by different service providers (or the same operator). The P D G s of W L A N , W i M a x , satellite and 3 G net-works are connected to different P - C S C F servers in IMS. In general, each A N has its own separate W A G . In addition, there are separate S - C S C F and I - C S C F servers for the two networks. For the establishment of an IMS session between two access networks, the two service providers should have a S L A with each other. The IMS networks which are owned 29 Chapter 3. LCSCW2 and TCSCW2: Architectures for IP Multimedia Subsystem Legend: I Data - Signaling PDN: Packet Data Network GGSN: Gateway GPRS Support Node SGSN: Serving GPRS Support Node BSC: Base Station Controller PDG: Packet Data Gateway IMS: IP Multimedia Subsystem RNC: Radio Network Controller WAG: Wireless Access Gateway PDSN: Packet Data Serving Node ARG: Access Router & Gateway WBSC: WiMax Base Station Controller - WNC: WiMax Network Controller HSS: Home Subscriber Server SFES: Satellite Fixed Earth Station ISL: Inter-Satellite Link SFES AAA: Authentication, Authorization and Accounting PSi Packet Switched UE: User Equipment AP: Access Point PCF: Packet Control Function S3GIF: Satellite-3G Interworking Function WMIF: WiMax-3G Interworking Function WIF: WLAN-3G Interworking Function 3GPP: Thjrd Generation Partnership Project SSB: Satellite Spot Beam Figure 3.2: The T C S C W 2 interworking architecture by different operators are connected together through an IMS backbone network. The P D G and P D S N are connected to the same P - C S C F server if W L A N , W i M a x , satellite and 3 G network are owned by the same operator. The mechanisms involved in the interworking architecture along with the functionali-ties of the important entities are explained below with reference to the 3 G P P Specification [31]. Access to external IP networks such as IMS, 3G operators network corporate In-tranets or the Internet through the 3 G P P system is called " W L A N 3 G P P IP Access". The P D G provides W L A N 3 G P P IP Access to external IP networks. In the T C S C W 2 architecture, a U E is identified by multiple IP addresses. For example, in case of a U E 30 Chapter 3. LCSCW2 and TCSCW2: Architectures for IP Multimedia Subsystem in W L A N accessing IMS or 3 G P P PS services, the U E is identified by two IP addresses i.e. a local IP address and a remote IP address. A local IP address is used to deliver a packet to the U E in W L A N A N . The local IP address identifies the U E in W L A N A N . The U E ' s local IP address may be translated by network address translation ( N A T ) before delivering the packet from U E to other IP network including public land mobile network ( P L M N ) . The remote IP address is used by the data packet encapsulated in-side the U E to P D G tunnel. The remote IP address identifies the U E in the network which the W L A N is accessing v ia P D G . A tunnel is established from the U E to P D G for carrying PS based services traffic in 3 G P P IP Access. The data for more than one IP flow and for different services may be carried in one tunnel. It may not be possible to separate individual IP flows and service traffic at intermediate nodes because of the possible encryption of the data including IP header within these tunnels. However, QoS can be assured if the W L A N U E and P D G deploy DS mechanism and appropriately color the DS field in the external IP header according to the QoS requirement of a particular traffic flow. The P D G assigns remote IP address to the W L A N U E . It registers the W L A N U E ' s local IP address and binds the U E local IP address wi th the U E remote IP address. The P D G also performs the encapsulation and decapsulation of packets since it is the terminating/originating point of tunnel between U E and P D G . The W A G performs collection of per tunnel accounting information e.g. byte count, elapsed time etc. and sends this charging information to the 3 G P P A A A server [31]. In the T C S C W 2 architecture, the W L A N - 3 G interworking function (WIF) , which is 31 Chapter 3. LCSCW2 and TCSCW2: Architectures for IP Multimedia Subsystem connected to the S G S N or P C F of the 3G core network, is responsible for hiding the details of the W L A N from the 3 G core network and implementation of the 3 G protocols for mobility management, authentication etc. essential for the 3 G radio access network. The W I F gives W L A N appearance of another 3 G A N from the perspective of 3 G core network. In other words, the W L A N is considered like another G P R S R A in the system. In the T C S C W 2 architecture, the satellite access network comprises of satellites and S F E S . Satellites convey data and signaling messages exchanged between U E and S F E S . S F E S performs the power control, link control, radio bearer control and paging functions. Satellites can be interconnected in an orbit v ia ISL. S F E S is connected to W A G for accessing 3 G P P P S and IMS services via P D G . The satellite-3G interworking function (S3GIF) is mainly responsible for connecting satellite systems wi th core 3G network. The S3GIF, which is connected to the S G S N / P C F of the U M T S / C D M A 2 0 0 0 core network, is responsible for hiding the details of the satellite network from the 3G core network. It is also responsible for the conversion of signaling and packet formats of satellite network to U M T S / C D M A 2 0 0 0 network and vice versa. The W i M a x A N consists of W i M a x base stations controlled by the W B S C . Many W B S C s are controlled by one W N C . The W N C is connected to W A G to provide W i M a x users with 3 G P P PS and IMS services via P D G . The W i M a x - 3 G interworking function ( W M I F ) connects the W N C to the core 3G network. The T C S C W 2 interworking architecture integrate satellite network, 3G, W i M a x , and W L A N based on the tight coupling approach since the satellite network, 3G , W i M a x , and W L A N are directly coupled to the 3 G network via interworking functions. 32 Chapter 3. LCSCW2 and TCSCW2: Architectures for IP Multimedia Subsystem For seamless operation in the T C S C W 2 architecture, U E s are required to implement the 3 G protocol stack on the top of their standard network cards. Among the disadvan-tages of tightly-coupled approach is exposure of the 3 G core network interfaces directly to the W L A N , W i M a x and satellite network which invites security challenges. Extensive efforts are required for the implementation of interworking functions especially for the A N s not owned by the 3 G operators. The 3 G core network entities i.e. S G S N and G G S N need to be modified to handle the increased load caused by the direct injection of the traffic from other A N s . The T C S C W 2 mandates the use of 3G-specific authentication mechanisms based on universal subscriber identity module (USIM) or removable-user identity module ( R - U I M ) cards for authentication in other A N s . This requires A N s to interconnect to the 3 G carriers' SS7 network for performing authentication procedures. Hence, either other A N s interface cards, for e.g., I E E E 802.11 W L A N network interface card, be equipped wi th buil t - in U S I M or R - U I M slots or external U S I M or R - U I M cards need to be plugged separately into the U E s . Among the advantages of the T C S C W 2 architecture is the possibility of reuse of A A A , mobility management and QoS handling infrastructures of 3 G cellular networks. The T C S C W 2 architecture enables the provision of 3 G services to other A N s users with guaranteed QoS and seamless mobility. How-ever, the 3 G core network nodes can not accommodate the bulky data traffic from the other A N s during busy hours since the core network nodes are designed to support the small-sized data of circuit voice calls or short packets. 33 Chapter 3. LCSCW2 and TCSCW2: Architectures for IP Multimedia Subsystem 3.3 An Analytical Model for Cost Analysis W e use the inter-system c o m m u n i c a t i o n cost ana lys is to evaluate the per fo rmance of our p roposed L C S C W 2 in t e rwork ing arch i tec ture . T h e t o t a l inter-system c o m m u n i c a t i o n cost Cs is g iven by: Cs = Ct + Cp + Cq (3.1) where Ct, Cp, a n d Cq denote the t r ansmiss ion cost, process ing cost, a n d queue ing cost, respect ively. T h e t r ansmiss ion cost Ct is the cost i n cu r r ed due to the t r ansmiss ion of s igna l ing and/o r da t a . It depends on the packet a r r i va l rate , the t r ansmiss ion ra te of the l i nk , a n d the d is tance between the ne ighbor ing network ent i t ies. T h e process ing cost Cp is the cost assoc iated w i t h the encapsu la t ion , decapsu la t ion and r o u t i n g of packets. T h e queue ing cost Cq is the cost i n cu r red due to the queu ing of packets i n each network entity. O u r ana lys is is app l i cab le t o b o t h I P v 4 a n d I P v 6 packet types. A l s o , our ana lys is is v a l i d for b o t h U M T S and C D M A 2 0 0 0 3 G networks. In the L C S C W 2 arch i tec ture , c o m m u n i c a t i o n pa ths are different when the source node (SN) resides i n ei ther W i M a x , W L A N or sate l l i te network. In the fo l low ing ana lys is , we assume tha t S N is c o m m u n i c a t i n g v i a a W L A N a n d the correspondent node ( C N ) is us ing a 3 G wireless ce l lu la r network. However, the analys is can easi ly be ex tended for the case when the S N is c o m m u n i c a t i n g v i a W i M a x or sate l l i te network. 34 Chapter 3. LCSCW2 and TCSCW2: Architectures for IP Multimedia Subsystem Originating Originating Terminating Terminating Visited Home Home Visited Network Network Network Network ! n !, H , SN A P J 132. 200 OK| « 33. ACK 31 200 O K U 3. UAR * 4. UAA #J 35 W A G Inter-net 36. A C K P-C S C F s-C S C F C S C F H..,-0.-L.'-RJ |26. 200 O K L -27. 200 O K | ^ . . . . . . V H H":---V-<a»-IMVITE„ I N V I T E P-C S C F 0 o ^ . 2 0 0 o £ - . ? ? ? ° * - 2 0 0 O K | H-4°:-^y-iACK G G S N S G S N !M.^ .*(:liJ43.ACK RNC £ 18, INVITEj bo. 200 OK i i ' ? ? ? °3 B S C 45. ACK G G S N : Gateway G P R S Support Node W A G : Wireless A c c e s s Gateway Data Signaling S N : Source Node A R G : A c c e s s Router & Gateway S G S N : Serving G P R S Support Node HSS: Home Subscr iber Server A A A : Authentication, Authorization and Account ing R N C : Radio Network Controller A P : A c c e s s Point B S C : Base Station Controller C N : Correpondent Node U A R : Diameter User-Authentication-Request LIR: Diameter Location-Information-Request U A A : Diameter User-Authentication-Answer L I A : Diameter Location-information-Answer P/S/ I -CSCF: Proxy/Serving/lnterrogating-Cail Sess ion Control Function Figure 3.3: Signaling and data communication paths between W L A N and 3 G for IMS session in the L C S C W 2 architecture [32], [25]. 3.3.1 Available Paths for Communications Figure 3.3 and 3.4 show the communication path between W L A N and 3 G for an IMS session in the L C S C W 2 and T C S C W 2 architecture, respectively. The solid arrows show the data traffic communication path whereas the dashed arrows indicate the signaling traffic communication path from the S N to the C N . We consider the IMS session estab-lishment signaling [32], [25] where the S N sends a SIP I N V I T E message to the C N . The SIP 200 O K message is sent from the C N to the S N . The SIP A C K message from the S N to the C N indicates the completion of session establishment procedure. The signaling 35 Chapter 3. LCSCW2 and TCSCW2: Architectures for IP Multimedia Subsystem Originating Originating Terminating Terminating Visited Home Home Visited Network Network Network Network I II 1 I II 1 SN 132.200 OKJ « 33. ACK 31. 200 OK -» •L?.-.*?.KJ UAR * 4. UAA p C S C F S-C S C F I-C S C F £6. 200 OK 1^ H S S 10. UR • 11. LiA S-C S C F P-C S C F G G S N - i3.iNvrrE, 4 I N V I T E ^ 15. 24.200 OK 2J;.J?.tl°K * |2i. M ° ° K | RNC B S C 18. INVITEJ CN ff:*C.Jj«.ACK G G S N : Gateway G P R S Support Node W A G : Wireless A c c e s s Gateway Signaling. S N : Source Node A R G : A c c e s s Router & Gateway S G S N : Serving G P R S Support Node H S S : Home Subscr iber Server A A A : Authentication, Authorization and Account ing R N C : Radio Network Controller A P : A c c e s s Point B S C : Base Station Controller C N : Correpondent Node P D G ; Packet Data Gateway UAR: Diameter User-Authentication-Request LIR: Diameter Location-Information-Request U A A : Diameter User-Authentication-Answer L I A : Diameter Location-Information-Answer P /S f l -CSCF: Proxy/Serving/lnterrogatlng-Call Sess ion Control Function Figure 3.4: Signaling and data communication paths between W L A N and 3G for IMS session in the T C S C W 2 architecture [32], [25]. incorporates the authentication from the 3 G P P A A A server and the query of the user's profile from the HSS database based on Diameter protocol messages [33]. Note that for simplicity, we do not consider the provisional responses such as "100 Trying" in the signaling path. 3.3.2 Transmission Cost Let A denote the I M S session arrival rate (requests per second) and 1 denote the number of packets per request both for signaling and data from a S N . For transmission and 36 Chapter 3. LCSCW2 and TCSCW2: Architectures for IP Multimedia Subsystem processing cost calculation of IMS signaling traffic, 1 is equal to 1 because we assume that one signaling packet can carry one particular signaling message such as 200 O K from one node to its adjacent node. We take into account the traffic coming from other users in the same A N as well as from other A N s by considering background util ization at network entities. The transmission cost between W L A N and 3G wireless cellular networks for IMS signaling traffic C f 9 is: A£(2</9 + ip(3dap—arg -{- 2darg—aaa - j - 2>darg—wag -\- 3dwag—inet ~\~ 3d j n e t_p C S C j "I - Gdpcscf—scscf -f- 4^dscscf^icscf + dscscf—scscf + 2,diCscf—flss -\- 3dpCSCf^.ggSn + 3dSgSn — ggSn "f" 3dggSn — r n c ~\~ drnC—l}SCj) (3-2) where cp and ip are the unit packet transmission costs in wireless and wired link respec-tively, dap—arg, darg—aaa, darg—wag, dwag—ineti dinet—pCSCf} dpCSCj—scscf, dscscf—icscf, dscscf—scscfi dicscf—hssi dpcscf—ggsm dSgSn—ggSn, dSgSn—rnc, and drnc—f)SC denote the distance between A P and A R G , A R G and A A A , A R G and W A G , W A G and Internet, Internet and P - C S C F , P - C S C F and S - C S C F , S - C S C F and I - C S C F , S - C S C F server of the S N IMS network and the S - C S C F server of the C N IMS network, I - C S C F and HSS, P - C S C F and G G S N , S G S N and G G S N , S G S N and R N C , and R N C and B S C , respectively. The distance is defined as the number of hops that a packet has traveled [34]. The transmission cost between W L A N and 3 G wireless cellular networks for IMS data 37 Chapter 3. LCSCW2 and TCSCW2: Architectures for IP Multimedia Subsystem traffic Cfata is: C\ = A£(2<£> + ijj(dap—arg -\- d a r g ^ . w a g -\~ dwag—inei + dinei—pCSCf + 2dpCSCf ^scscj ~f~ dsCSCf — SCSCf dpCSCJ—ggSn ~\~ dggSn — Sggn ~\~ dggSn — r n C ~\~ drnC—(>SC) ) (3.3) By following the same methodology, we can calculate the transmission cost for commu-nication paths between WiMax and 3 G , as well as satellite and 3 G for IMS traffic. 3.3.3 Processing Cost For processing cost calculation, we first assume that Nbsc BSCs are connected to each RNC, Nrnc RNCs are connected to each SGSN, Nsgsn SGSNs are connected to each GGSN, NggSn GGSNs and Npcscf P-CSCFs are connected to the Internet. In addition, let Nmni, Nmn2, Nmn3, and Nmn4 denote the number of users in the coverage area of 3 G wireless cellular network, W L A N , WiMax, and SSB of the satellite network, respectively. The total number of users N in the network can be given as: N = Nmnl + Nmn2 + Nmn3 + Nmn4 (3.4) The processing cost between W L A N and 3G wireless cellular network for IMS signaling traffic C*3 is: Cp 9 — 3Cp—ap + 4 C p _ a r g -f" Cp—aaa "f" 3Cp—wag -\- 3Cp—{nei + QCp—pCSCf -\~ QCp—Scscf ~t~~ 3Cp—iCSCf -f~ Cp—}igS -\~ 3Cp—ggSn -f- 3Cp—SgSn ~\~ 3C1p—mc "F 3Cp—bsc (3.5) where C p _ a p represents the processing cost at A P and is given as: Cp-ap = ^-lap (3-6) Chapter 3. LCSCW2 and TCSCW2: Architectures for IP Multimedia Subsystem where 7 a p denotes the unit packet processing cost at A P . The unit packet processing Cp—wagi Cp—pCSCf, Cp^scscfi and Cp—icscj represent the processing costs at A R G , W A G , P -C S C F , S - C S C F and I - C S C F , respectively. Their expressions are similar to that of C p - a p with the only difference that they have their own respective unit packet processing costs. Cp-aaa represents the processing cost at the A A A server and is given by: where ^ a a a denotes the unit packet processing cost at A A A server. We assume that IP addresses are searched in the lookup table using the multiway and multicolumn search [35]. We also assume that the number of entries in the lookup tables for A A A server and HSS are equal to the total number of users iV in the network because 3 G P P A A A server based authentication and subscription database HSS are used [31]. In addition, L is the IP address length in bits (e.g. L is 32 for IPv4 and 128 for IPv6), S is the machine word size in bits, and k is a system-dependent constant. In our analysis, u>i where i G {1,2 ,3 ,4} denotes the weighting factors. Cp-hss represents the processing cost at HSS and its expression is similar to that of C p _ a a a wi th the only difference that it has its own specific unit packet processing cost. Cp-ggsn, C p - s g s r u C p - r n c , and Cp-bsc represents the processing costs at G G S N , S G S N , R N C , and B S C respectively wi th similar expressions as that of C p - a a a wi th the difference that they have their own respective unit packet processing costs. Also, the logarithm is taken for N s g s n in case of C p - g g s n , N r n c in case of C p _ s g s n , Nbsc in case of C p - r n c , and i V m n l in case of C p _ ( , s c instead of N in the cost includes the cost for encapsulation and decapsulation of packets. Similarly, C} 'p—arg) (3.7) 39 Chapter 3. LCSCW2 and TCSCW2: Architectures for IP Multimedia Subsystem expression of Cp-aaa. Cp-inet represents the processing cost at the Internet and is given as: Cp-inet = U (^inet + u2 [\ogk+l(Ngp) + ) (3-8) where 7 i n e t denotes the unit packet processing cost at the Internet, N g p = N g g s n + Npcscf, and £ is equal to 1 for IMS signaling processing cost calculation. The processing cost between WLAN and 3G wireless cellular network for IMS data traffic C^ata is: Cp Cp—ap + Cp—arg + Cp—yjag H~ Cp—inet ~f" ^Cp—pcscf "I- 2 C p _ s c s c j -f" Cp- •ggsn ~t~ Cp—sgsn -)- Cp—rnc -\~ Cp—i)SC (3.9) Following the same approach, we can calculate the processing cost for communication paths between WiMax and 3G, as well as satellite and 3G for IMS traffic. 3.3.4 Queueing Cost For the queueing cost calculation, we first model the communication path between SN and CN as a network of M / M / l queues [36], [37]. The queueing cost is proportional to the total number of packets in the queueing network. The queueing cost between WLAN and 3G wireless cellular network for IMS signaling traffic Csq19 is: Csq19 = u3(3E[nap] + 4E[narg] + E[naaa] + 3E[nwag] + 3E[ninet] + 6E[npcscf] + 6E[nscscf] + 3E[nicscf] + E[nhss] + 3E[nggsn] + 3E[nsgsn] + 3E[nrnc] + 3E[nbsc]) (3.10) 40 Chapter 3. LCSCW2 and TCSCW2: Architectures for IP Multimedia Subsystem where E[nap], E[narg], E[naaa], E[nwag], E[ninet], E[npcscf], E[nscscf] E[nicscf], E[nhss], E[nggsn], E[nsgsn], E[nrnc], E[nbsc] denote the expected number of packets in the queue of A P , A R G , A A A , W A G , Internet, P - C S C F , S - C S C F , I - C S C F , HSS, G G S N , S G S N , R N C , and B S C , respectively. The value of E[nap] is equal to: E[nap] = (3.11) J- Pap where pap = A e _ a p / / i a p represents the util ization at A P queue, \iap denotes the service rate at A P queue and \ e - a p represents the effective arrival rate (in packets per second) at A P queue. That is, A e _ a p = J2ieNap A i , where Nap denotes the number of active users in the A P coverage area that are engaged in communication with the A P , and hence Nap Q Nmn2. The effective arrival rate A e at a network node can be determined from the util ization at that node. Similarly, the A e at queues of other network nodes can be calculated and expressions can be determined for the expected number of packets at other network entities. The queueing cost between W L A N and 3 G wireless cellular network for IMS data traffic C j ° t a is: Cgata = u4(E[nap] + E[narg] + E[nwag] + E[ninet] + 2E[npcscf] + 2E[nscscf] + E[nggsn] + E[nsgsn] + E[nrnc] + E[nbsc\) (3.12) Following the same approach, we can calculate the queueing cost for communication paths between W i M a x and 3G, as well as satellite and 3 G for IMS traffic. 41 Chapter 3. LCSCW2 and TCSCW2: Architectures for IP Multimedia Subsystem 3.4 Numerical Results In this section, we present the numerical results for the cost analysis of our proposed L C S C W 2 and T C S C W 2 interworking architectures. The total system signaling and data costs for IMS traffic are determined for the case when the S N is using the W L A N and the C N is in 3 G wireless cellular network. We consider a network containing two 3 G BSCs , three W i M a x B S C s , 12 W L A N s , and one SSB. The cell radius for 3 G B S C , W i M a x B S C , and W L A N is taken to be 1000 m, 700 m, and 50 m, respectively. Their user densities are taken to be 0.001, 0.001, and 0.008 per square meter, respectively [30], [29], [38]. The SSB is assumed to cover an area of 20 square kilometer and user density in its coverage area is taken to be 0.0005 per square meter [12]. The number of users resulting from the selection of these cell radii and user densities in different A N s are: Nmnl = 5000, Nmn2 = 600, Nmn3 = 3000, and Nmn4 = 10000. In our network setting, two G G S N s and two P - C S C F servers are connected to the Internet; each G G S N supports three SGSNs; each S G S N supports four R N C s , and each R N C controls five BSCs . The IP address length L and processor machine word size S are taken to be 32 bits. The system dependent constant value k is equal to 5 [35]. The wired hop distances, dpcscf-ggsn and dsgsn-rnc, which involve the core 3 G network entities are equal to 4, and rest of the distances are equal to 2 [30], [29], [39]. The trunked Pareto distribution is assumed for packet length with average packet length equal to 480 bytes. The inter-arrival time for packets is exponentially distributed [40]. The weighting factors, LOI and u>2, corresponding to the table lookup processing cost 42 Chapter 3. LCSCW2 and TCSCW2: Architectures for IP Multimedia Subsystem are taken equal to 1 x 10~ 6 as lookup delay is increased by 100 ns for each memory access [35]. The weighting factors, u3 and corresponding to queueing cost are equal to each other and are chosen such that sum of all the weighting factors is equal to 1 (i.e. = 1). We consider wireless link channels to be 9.6 kbps, 19.2 kbps or 19.2 kbps and the wired links to be 1 Gbps. The unit transmission costs for the wired link ip and the wireless link ip are equal to 3.84 x 10~ 6 and 0.1, respectively [41], [42] so that the unit transmission costs can be interpreted as typical wireless and wired link delays in seconds. The service rate p, at al l the network entities is taken equal to 250 packets/sec. The unit packet processing cost for al l the network entities is taken equal to 4 x 10~ 3 except the core 3 G network entities i.e. S G S N and G G S N and the Internet for which the unit packet processing cost is taken twice as compared to other network entities in accordance wi th [30], [29]. For IMS data traffic, we consider audio and video sessions using different codecs which give different packet generation rates. For instance, G S M voice encoder at 13 kbps, G.726 voice encoder at 32 kbps, H . 2 6 4 / A V C at 56 kbps, H . 2 6 4 / A V C at 80 kbps, H . 2 6 4 / A V C at 90 kbps give packet generation rates of 4, 9, 15, and 21 packets/sec, respectively [43], [19]. The background uti l ization due to traffic from other sources is taken to be 0.7 for HSS, A A A server, and Internet because they have to handle traffic for inter-system communications from different A N s , 0.5 for the core 3 G entities i.e. S G S N and G G S N , and 0.4 for the rest of the entities. The assumption of these values of background utilizations allows us to determine A e at each of the network nodes. 43 Chapter 3. LCSCW2 and TCSCW2: Architectures for IP Multimedia Subsystem Figure 3.5: The breakup of system signaling cost into transmission, processing and queue-ing cost for different values of IMS signaling arrival rate A in the L C S C W 2 architecture. Figure 3.5 shows the transmission, processing and queueing costs, as well as the total system signaling cost, for IMS signaling traffic in the L C S C W 2 architecture. Results show that that the ratio Ct : Cv is 1 : 1.02 for signaling in the L C S C W 2 architecture. W i t h our selection of parameters, queueing cost is higher than the transmission and processing costs for signaling and lower than the transmission and processing costs for data. Figure 3.6 shows the transmission, processing and queueing costs, as well as the total system signaling cost, for IMS signaling traffic in the T C S C W 2 architecture. Results show that that the ratio Ct : Cp is 1 : 0.959 for signaling in the T C S C W 2 architecture. W i t h our selection of parameters, queueing cost is higher than the transmission and processing costs for signaling and lower than the transmission and processing costs for 35 IMS Signaling Arrival Rate k (packets per second) 44 Chapter 3. LCSCW2 and TCSCW2: Architectures for IP Multimedia Subsystem 30 IMS Signaling Arrival Rate X (packets per second) Figure 3.6: The breakup of system signaling cost into transmission, processing and queue-ing cost for different values of IMS signaling arrival rate A in the T C S C W 2 architecture. data. Figure 3.7 shows the effect of varying IMS signaling arrival rate A on total system signaling cost in the L C S C W 2 and T C S C W 2 architectures. It can be observed that the system signaling cost increases almost linearly with the increasing value of A in both the architectures. A comparison of the system signaling cost in the L C S C W 2 and T C S C W 2 architectures reveals that the signaling cost in the T C S C W 2 architecture is always considerably less than the L C S C W 2 architecture for all values of A. A reduction in the system signaling cost is an important achievement of the T C S C W 2 architecture which justifies the deployment of P D G s . It is to be noted that signaling cost is most cri t ical in the networks because session establishment, session resource reservation, U E registration, 45 Chapter 3. LCSCW2 and TCSCW2: Architectures for IP Multimedia Subsystem 40 15' 1 1 ' ' 1 0 5 10 15 20 25 IMS Signaling Arrival Rate X (packets per second) Figure 3.7: Effect of varying the IMS signaling arrival rate A on the total system signaling cost between W L A N and 3G in the LCSCW2 and TCSCW2 architectures. and vertical handoffs are achieved via the signaling. Hence, QoS can be guaranteed in the TCSCW2 architecture. In the TCSCW2 architecture, each A N has its own P D G via which the traffic is routed to IMS network without passing through the Internet and admission control mechanisms can be easily implemented to guarantee the QoS. In the LCSCW2 architecture, IMS network is reached v ia Internet whose background util ization can vary at different times giving an almost best effort service in the LCSCW2 architecture. Figure 3.8 shows the effect of varying IMS data traffic arrival rate A resulting from using different audio and video encoders on total system data cost in the LCSCW2 and TCSCW2 architectures. It can be observed that the system data cost increases non-46 Chapter 3. LCSCW2 and TCSCW2: Architectures for IP Multimedia Subsystem 5000 IMS Data Arrival Rate X (packets per second) Figure 3.8: Effect of varying the I M S data arrival rate A on the total system data cost between W L A N and 3G in the L C S C W 2 and T C S C W 2 architectures. linearly with the increasing value of A in both the architectures. It can be noticed that the data cost is almost the same in both the architectures for our considered parameters. This observation implies that for data traffic both the architectures are able to provide similar k ind of service. The linear increase of system signaling cost and non-linear increase of system data cost with A is dependent on the ratios of transmission, processing and queueing costs in the total system cost. Figure 3.9 shows the effect of varying IMS session duration on total system data cost in the L C S C W 2 and T C S C W 2 architectures. The arrival rate A is assumed to be 21 packets/sec. The IMS data session is run for 30, 60, 120, 240, and 480 seconds with the corresponding values of £ as 315, 630, 1260, 2520, 5040, and 10080, respectively. 47 Chapter 3. LCSCW2 and TCSCW2: Architectures for IP Multimedia Subsystem °0 50 100 150 200 250 300 350 400 450 500 IMS Data Session Duration (seconds) Figure 3.9: Effect of varying the average IMS session duration I on the total system data cost between W L A N and 3 G in the L C S C W 2 and T C S C W 2 architectures. Results show that the system data cost increases almost linearly wi th increasing session length for both the architectures and the system data cost is almost the same in the two architectures. 3.5 Summary In this chapter, we proposed the L C S C W 2 and T C S C W 2 interworking architectures for 4G heterogeneous wireless networks. The L C S C W 2 and T C S C W 2 architectures inte-grates the satellite networks, 3 G wireless networks, W i M a x , and W L A N s . The L C S C W 2 architecture supports I M S sessions, provides global coverage, and facilitates independent deployment of various access networks. We also proposed a cost model to determine the 48 Chapter 3. LCSCW2 and TCSCW2: Architectures for IP Multimedia Subsystem associate cost for the IMS signaling and data traffic in the LCSCW2 architecture. We presented the numerical results for the system cost, as well as the transmission, process-ing, and queueing costs under different arrival rates and session lengths for LCSCW2 and TCSCW2 architectures. 4 9 Chapter 4 Analysis of SIP-based Signaling for I M S Sessions in 4 G Networks In this chapter, we study the SIP-based IMS signaling delay for registration, session es-tablishment and session re-establishment for different 3G and W L A N channel rates. The S N and the C N needs to register themselves wi th the IMS network before an IMS session can be established between them. When the U E (SN or C N ) moves to a new network during an IMS session which is referred to as "mid-session mobili ty", the terminal needs to inform the other node by sending a SIP r e - I N V I T E message containing information about the terminals' new IP address and to update the session description. However, during the mid-session mobility, before a terminal can send a r e - I N V I T E message to the other node, a terminal requires to perform some procedures for attaching itself to the new A N infrastructure. For instance, a terminal attaches to 3G U M T S network using G P R S attach and P D P context activation procedure. A terminal uses D H C P registration procedure for connecting to the A N of W L A N . We calculate the delay for I M S session re-establishment after undergoing a vertical handoff for different scenarios. 50 Chapter 4. Analysis of SIP-based Signaling for IMS Sessions in 4G Networks 4 . 1 Delay Analysis of IMS Signaling Procedures In this section, we analyze the delay for the IMS signaling procedures. The delay consists of three parts, i.e., transmission delay, processing delay, and queueing delay. Dtotal Dtrans "f" Dproc ~f" Dqueue where D t o t a i denotes the total delay for a signaling procedure, D t r a n s denotes the trans-mission delay, D p r o c denotes the processing delay and D q u e u e denotes the queueing delay. The transmission delay considered here includes the propagation delay as well. The transmission delay is the delay incurred due to the transmission of signaling which de-pends on the message sizes as well as the rate or the bandwidth of the channel and the delay incurred due to the propagation of signaling messages from one node to another which depends on the distance between the nodes. The processing delay is the delay associated with the encapsulation, decapsulation and routing of packets. The queueing delay is the delay incurred due to the queuing of packets in the queues at each node. Our analysis is equally valid for IPv4 as well as IPv6. Depending on the A N s in 4 G wireless networks, there wi l l be different entities in the path between S N or C N and the P - C S C F server of the I M S . Hence transmission, processing and queuing delays of the intervening components wi l l be added to the total delay. For instance, if the S N is in 3G U M T S network, then the path between S N and the P - C S C F server in the L C S C S W 2 and T C S C W 2 interworking architectures is given as: SN -> BSC -> RNC SGSN GGSN P - CSCF (4.2) 51 Chapter 4. Analysis of SIP-based Signaling for IMS Sessions in 4G Networks If the S N is in W L A N , then the path between S N and the P - C S C F server in the L C S C W 2 architecture is given as: SN AP -> ARG -> WAG -> Internet -> P - CSCF (4.3) If the S N is in W L A N , then the path between S N and the P - C S C F server in the T C S C W 2 architecture is given as: SN -> AP ARG -»• WAG ->• P D G -»• P - CSCF (4.4) For the G P R S attach and packet data protocol ( P D P ) context activation procedure, the B S C and R N C are there between S N and S G S N but we do not consider them in our analysis for simplicity. We do not consider the A N entities between the S N or C N and the P - C S C F for the IMS registration and session establishment procedure to make our analysis independent of any particular A N . We consider wireless link transmission from S N or C N to the P - C S C F server directly which mainly contributes to the delay. Hence, our assumptions for the simplification of delay analysis have negligible effect on the results. The considered assumptions can be easily relaxed to obtain the exact delay results corresponding to different A N s . 4.1.1 Transmission Delay We only consider wireless link transmission delays as the wired link transmission delays between the core network entities can be considered to be negligible because of the high bandwidths in the wired links. For wireless link transmission, we use the analytical 52 Chapter 4. Analysis of SIP-based Signaling for IMS Sessions in 4G Networks model considered in [41]. The wireless link transmission is analytically modeled with and without R L P with T C P as transport layer protocol. Average delay for receiving a packet (e.g. SIP I N V I T E packet) when no R L P is used is given as [41]: " 2 - {l-p)K1 D a v g n o R L P — T + D + (K-1)T (4.5) 2(1 -p)K J Here p is the probability of a frame being in error; K is the number of frames per packet; D is the end-to-end frame propagation delay over the radio channel; r is the inter-frame time; T is the packet transmission interval. The average delay to receive a packet when R L P is used is given by [41]: D a v g R L P — T + D R L P (4.6) 2P* Here DRLP denotes the delay experienced by a packet when R L P is used and is given as [41]: K(Ps-(l-p)) D R L P = D + (K-1)T + P2 s Z Z P & ) ( y D + ( M ± 2 . + *),)] (4.7) where Ps indicates the probability of transmitting a frame successfully over the R L P given by [41]: Ps = l-p(p(2-p))^ (4.8) The Sij denotes the first frame received successfully at the destination, the frame is the ith retransmission frame at the jth retransmission tr ial , and its probability is given as [41]: P(5y) = p(l - p)2 ((2 - p)p)m^1+l-1 (4.9) 53 Chapter 4. Analysis of SIP-based Signaling for IMS Sessions in 4G Networks The average delay for successfully transmitting a T C P segment wi th no more than N m a x T C P retransmission trials without R L P operating underneath is given as [41]: D, a v g T C P n o R L P {K-1)T + , I " ? D I - qNmaXTCp)(l - 2q) 1 - q N„ -D Q 1 - q i T C P + l qNmaxTCP 1 - 2q (4-10) where NmaxTcp indicates the maximum allowable T C P retransmissions, and q denotes the packet loss rate without R L P and is given as [41]: q = \ - ( l - p ) K (4.11) The average delay for successfully transmitting a T C P segment wi th no more than N m a x T C P retransmission trials wi th R L P operating underneath is given as [41]: D, a v g T C P w i t h R L P = DRLP + 2£>r(l. - r) 1 — fNmaxTCP 4r ( l - (2r N„ ') r ( l -1 - 2r 1 - r (4.12) where r denotes the packet loss rate wi th R L P and is given as [41]: r = l - ( l - p ( p ( 2 - p ) ) 6\K (4.13) The maximum sizes of the IMS signaling messages have been selected and hence, our results wi l l be conservative and wi l l give an upper bound on the analyzed delays. The maximum size of the messages exchanged in the G P R S attach procedure is 43 bytes, P D P context activation procedure is 537 bytes, D H C P registration procedure is 548 bytes, and Diameter authentication messages is 40 bytes [23, 33, 44, 45]. For SIP messages, SigComp 54 Chapter 4. Analysis of SIP-based Signaling for IMS Sessions in 4G Networks Table 4.1: Size of SIP messages involved in IMS signaling SIP Message Type Compressed Size (with SigComp) I N V I T E 810 R E G I S T E R 225 183 S E S S I O N P R O G R E S S 260 P R A C K 260 100 T R Y I N G 260 180 R I N G I N G 260 200 O K 100 A C K 60 S U B S C R I B E 100 N O T I F Y 100 401 U N A U T H O R I Z E D 100 U P D A T E 260 has been used which was developed by I E T F for compression of general text-based pro-tocols [46], [47], [48], [49]. It has been shown in [50] that S I P / S D P message sizes can be reduced by as much as 88% using SigComp wi th negligible compression/decompression time. The compression rate for the ini t ia l SIP messages such as I N V I T E , R E G I S T E R has been chosen to be 55% and for the subsequent SIP messages to be 80%. The SIP message sizes that have been selected according to standards [20, 51] are shown in Table 4.1. 55 Chapter 4. Analysis of SIP-based Signaling for IMS Sessions in 4G Networks We consider 9.6 kbps, 19.2 kbps, 128 kbps channel for 3G network; and 2 Mbps, 11 Mbps channel for I E E E 802.11 W L A N . We need to calculate the values of K i.e. the number of frames in a packet, for the above mentioned channel rates. The R L P frame duration or inter-frame time r is assumed to be 20 ms for 3 G A N [41]. We proceed as follows for calculation of value of K for 9.6 kbps channel: Number of bytes in each frame 9.6 x 10 3 x 20 x 10~ 3 x | = 24 bytes. For G P R S attach procedure, and 9.6 kbps channel; the value of K comes out to be |~||] = 2. W L A N frame duration is assumed to be 3.5 ms and inter-frame time r is taken to be 1 ms and is independent of the channel bit rate [23]. For 2 Mbps channel: Number of bytes in each frame 2 x 10 6 x 3.5 x 10" 3 x \ = 875 bytes. For D H C P registration procedure, and 2 Mbps channel; the value of K comes out to be Tiff] = 1. The values of K obtained for different messages following the same methodology are shown in Table 4.2. We take care of the value of K for a particular signaling message in our analysis. For G P R S attach procedure, 9 message are exchanged between U E (SN or C N ) and the S G S N of the new U M T S network [44] as shown in Figure 4.1. R L P is used on the top of medium access control ( M A C ) layer in 3 G networks to improve the bit error rate ( B E R ) performance. Hence, transmission delay for G P R S attach procedure D t r Q n s _ 9 p r s a t t a c / l is given by: k ) t r a n s — g p r s a t t a c h 9 X D a v g R L P (4-14) For P D P context activation procedure, 2 message exchanges are there between U E and S G S N [44] as depicted in Figure 4.2. Therefore, transmission delay for P D P context 56 Chapter 4. Analysis of SIP-based Signaling for IMS Sessions in 4G Networks Table 4.2: Values of K for signaling messages in different channel rates Message Types 9.6 kbps 19.2 kbps 128 kbps 2 Mbps 11 Mbps G P R S attach 2 1 1 1 1 P D P context activation 23 12 2 1 1 D H C P registration 23 12 2 1 1 SIP I N V I T E 34 17 3 1 1 SIP R E G I S T E R 10 5 1 1 1 183 S E S S I O N P R O G R E S S 11 6 1 1 1 SIP 180 R I N G I N G 11 6 1 1 1 SIP P R A C K 11 6 1 1 1 SIP 100 T R Y I N G 11 6 1 1 1 SIP U P D A T E 11 6 1 1 1 SIP 200 O K 5 3 1 1 1 SIP S U B S C R I B E 5 3 1 1 1 SIP N O T I F Y 5 3 1 1 1 SIP 401 U N A U T H O R I Z E D 5 3 1 1 1 SIP A C K 3 2 1 1 1 57 Chapter 4. Analysis of SIP-based Signaling for IMS Sessions in 4G Networks UE . GMM Attaci Request {TMSI, MNC, MCCT LAC, R A C J _ _ Search for TMSI 4. Identity Request 5. Identity Response 6. Authentication^ Request (RAND) Pass the RAND value to the SIM & obtain the Kc'and SRES yaluesi 7. Authentication Response ^ (SRES) Ji. Identity Check Request 9. Identity Check Response »' 18. GMM Attach Accept 19. GMM Attach Complete ^ 2. Identity Request (TMSI) 3. Identity Responle] !Msi) 10. IMEI 11. IMEI 15. Insert 16. Insert Check Request Check Response 13. Cancel 14. Cancel Subscriber Data Subscriber Data ACW 17. UpdatejLocation ACK H2. Update Location Location Location ACK UE: User Equipment SGSN: Serving GPRS Support Node EIR: Equipment Identification Register HLR: Home Location Register GMM: GPRS Mobility Management TMSI: Temporary Mobile Subscriber ID SRES: Signed Response MNC: Mobile Network Code MCC: Mobile Country Code LAC: Location Area Code RAC: Routing Area Code RAND: A Random Value SIM: Subscriber Identification Module IMEI: International Mobile Equipment Identity F igu re 4.1: G P R S a t t a ch procedure [44]. 58 Chapter 4. Analysis of SIP-based Signaling for IMS Sessions in 4G Networks activation procedure D t r a n s _ p d p c o n t e x t is given as: Dtrans—pdpcontext 2 X DavgRLP (4-15) For D H C P registration procedure, 4 message exchanges are there between U E and D H C P server [45] as shown in Figure 4.3. No R L P is used in case of W L A N because of much higher bandwidth and indoor operations. Therefore, D H C P registration procedure trans-mission delay D t r a n s _ d h c p is given by: Dtrans—dhcp = 4 X DavgTCPnoRLP (4-16) For IMS registration procedure including subscription to "reg Event" state, 8 message exchanges are there between U E and the P - C S C F server of the IMS network [32], [25] as shown in Figure 4.4. When registration procedure takes place in the 3 G A N , the transmission delay for IMS registrion procedure D t r a n s _ i m s r e g 3 g is given as: U trans—imsreg'ig — 8 X DavgTcpwithRLP (4-17) When registration procedure takes place in W L A N A N , the transmission delay for IMS registration procedure D t r a n s _ i m s r e g w l a n is given as: Dtrans—imsregwlan = 8 X DavgTCPnoRLP (4-18) For IMS session setup, 13 message exchanges are involved between S N and P - C S C F of the visited IMS network and 13 message exchanges are involved between P - C S C F of the terminating IMS network and C N [32], [25] as shown in Figure 4.5. When S N is in U M T S 59 Chapter 4. Analysis of SIP-based Signaling for IMS Sessions in 4G Networks network and C N is in W L A N and vice versa, the IMS session setup transmission delay D t r a n s — s e t u p — u w / w u is given by: - ^ t r a n s — s e t u p — u w / w u 12 X D a v g T C P w i t h R L P + 12 X D a v g T C P n o R L P (4-19) When S N as well as C N are in 3G network, the IMS session setup transmission delay D t r a n s — s e t u p — u u is given as. D t r a n s - s e t u p - u u = 24 X D a v g T C P w i t h R L P (4.20) When S N as well as C N reside in W L A N , the IMS session setup transmission delay P ^ t r a n s — s e t u p — w w is given by: D t r a n s - s e t u p - w w = 24 X D a V g T C P n o R L P (4-21) 4.1.2 Processing Delay We calculate the processing delay for different entities in the IMS signaling path. The processing delay for some of the nodes such as HSS, home location register ( H L R ) , and equipment identification register (EIR) mainly consists of the address lookup table delay. When a query is sent to HSS, H L R , or E I R for a particular IP address, the HSS, H L R , or E I R have to lookup their table for the given IP address. We assume that HSS, H L R , and E I R tables contains the list of all the users N in the network. The IP address lookup is the main component involved in the processing delay for databases. It has been shown that cache line size can be used to help in multiway search; and binary search can be adapted to perform multiple-column search for long length IP addresses [35]. For the 60 Chapter 4. Analysis of SIP-based Signaling for IMS Sessions in 4G Networks rest of the network entities, we assume a fixed processing delay mainly consisting of the delay involved in the encapsulation and decapsulation of packets. The processing delay in nanoseconds at the HSS, H L R or E I R can be approximated as: dproc—hss/hlr/eir — dproc-ed + 100 ^ l 0 g f c + 1 N + ^j ns (4.22) where L is the IP address length in bits e.g. L is 32 for IPv4 and 128 for IPv6, 5 is the machine word size in bits, k is a system-dependent constant, and dproc-ed represents the fixed processing delay due to the encapsulation and decapsulation of packets. We have used the multiplication factor of 100 ns in the above equation because it has been shown in [35] that the lookup time is increased by around 100 ns for each memory access. Considering the signaling flows in the G P R S attach procedure, the processing delay for the G P R S attach procedure Dproc-gprsattach can be given as: Dproc—gprsattach 4:dproc—sn -f- Qdproc—nSgSn -\- 2dproc—OSgSn -\- dproc—eir -\- 3<iproc_/[;r (4.23) where dproc—sn) dproc—nSgSn, dproc—OSgSn, dproc—eirt and dproc—/iir indicates a unit packet pro-cessing delay at S N , new S G S N , old S G S N , E I R and H L R , respectively. The processing delay for the P D P context activation procedure Dproc_p(ipcontext can be given as: Dproc—pdpcontext ^proc—sn ~f~ <$dproc—sgsn -\- dproc—dns -\- 3dproc—ggsn ~\~dproc—radius d~ dproc—dhcp (4.24) where dproc—SgSn, dproc—dnsi dproc—ggSn, dproc—radiusi and dproc—dhcp indicates a unit packet processing delay at S G S N , domain name system (DNS) server, G G S N , Radius server, 61 Chapter 4. Analysis of SIP-based Signaling for IMS Sessions in 4G Networks and D H C P server, respectively. The processing delay for the D H C P registration process Dproc-dhcp can be given as: Dproc—dhcp 2c?p r o c_3 n -)- 2,dproc—dhcp (4.25) The processing delay for the I M S registration process D p r o c _ i r n s r e g can be given as: Dproc—imsreg 4 d p r o c _ s n + Wdproc—pcscf + Qdproc—icscf + 4rfp r o c _/ l s s + 8 d p r o c — s c s c j (4.26) where d p r o c _ p c s c f , d p r o c _ i c s c f , d p r o c _ h s s , and d p r o c - s c s c f denotes the unit packet processing delay at P - C S C F , I - C S C F , HSS, and S - C S C F , respectively. The processing delay for the IMS session setup D p r o c _ i m s s e t U p can be given as: Dproc—setup 7dproc—sn "F 26(ip r o c_p C S Cj + 2,6dproc—scscf -\- 6dproc—icscf - \ ~ d p r Q c - h g g -\- Gdproc—cn (4.27) where d p r o c _ c n denotes the unit packet processing delay at the C N . 4.1.3 Queueing Delay We calculate the queueing delays for different network entities involved in the IMS sig-naling. The packet delay to reach from S N to C N depends on the queueing delay at each of the intervening queues which itself depends upon the number of packets at each queue. Also waiting time of a packet in a queue depends upon the number of packets in that queue. We have assumed M / M / l queues for the network entities. A queueing network is said to be in equilibrium if a stationary state exists. For an M / M / l queue in equilibrium 62 Chapter 4. Analysis of SIP-based Signaling for IMS Sessions in 4G Networks state, if the input process is Poisson wi th rate A, then the output process of the queue is also Poisson with rate A [37]. For a queueing network with M / M / l queues in tandem, if the input process to the first M / M / l queue is Poisson, the input process to the next stage M / M / 1 queue is also Poisson and independent of the input process and so on [36]. The expected total waiting time or delay in the queueing network consisting of queues in tandem is the sum of the expected waiting times at each queue. Considering the signaling flows in the G P R S attach procedure, the queueing delay for the G P R S attach procedure Dqueue-gprsattach can be given as: Dqueue—gprsattach ~ 4E[wsn] + 9E[wnsgsn] + 2E[wosgsn] + E[weir] + SE[whlr] (4.28) where E[wsn], E[wnsgsn], E[wosgsn], E[weir], and E[whir] indicates the expected value of a unit packet queueing delay at S N , new S G S N , old S G S N , E I R and H L R , respectively. The expected waiting time or delay of a packet at S N queue is given by [36]: E[wm] = (P°n_ , (4.29) Psn { J- Psn ) where psn = A e _ s n / p s n represents the uti l ization at S N queue, yu s n denotes the service rate at S N queue and A e _ s n represents the effective arrival rate (in packets per second) at S N queue. That is, A e _ s n = 2~2I<EN ^> w-h-ere Nsn denotes the number of active sessions apart from the considered IMS session at the S N . The effective arrival rate A e at a network node can be determined from the util ization at that node. Similarly, the A e at queues of other network nodes can be calculated and expressions can be determined for the expected waiting time at other network entities. The queueing delay for the P D P 63 Chapter 4. Analysis of SIP-based Signaling for IMS Sessions in 4G Networks context activation procedure D q u e u e - p d p c o n t e x t can be given as: E queue—pdpcontext E[wsn] +3E[wsgsn] + E[wdns] + 3E[wggsn] + E[wradius}+E[wdhcp} (4.30) where E[wsgsn], E[wdns], E[wggsn], E[wradius], and E[wdhcp] indicates the expected value of a unit packet queueing delay at S G S N , D N S server, G G S N , Radius server, and D H C P server, respectively. The queueing delay for the D H C P registration process Dqueue_dhcp can be given as: D queue—dhcp 2E[wsn] + 2E[wdhcp] (4.31) The queueing delay for the I M S registration process Dqueue_imsreg can be given as: D queue—imsreg = 4E[wsn] + 10E[wpcscf} + 6E[wlcscf] + AE[whss] + 8E[wscscf] (4.32) where E[wpcscf], E[iUiCSCf], E[wnss], and E[wscscf] denotes the expected value of a unit packet queueing delay at P - C S C F , I - C S C F , HSS, and S - C S C F , respectively. The queue-ing delay for the IMS session setup D q u e u e _ i m s s e t u p can be given as: E'queue—imssetup = 7E[wsn] + 26E[wpcscf] + 26E[wscscf] + 6E[wicscf] +E[whss] + 6E[wcn] (4.33) where E[wcn] denotes the expected value of a unit packet queueing delay at the C N . 4.1.4 Total Delay We calculate the total delay for IMS registration, session establishment, and session re-establishment. The delay for G P R S attach procedure is given as: Dgprsattach D^rans—gprsanac)l + Eproc—gprsaitach ~T" Dqueue_gprsattaca (4.34) 64 Chapter 4. Analysis of SIP-based Signaling for IMS Sessions in 4G Networks The delay for P D P context activation procedure is given by: Dpdpcontext = D trans—pdpcontext ~f~ Dproc—pdpcontext ~b D' queue—pdpcontext (4.35) The delay for D H C P registration procedure is given as: Ddhcp = Dtrans—dhcp ~r" Dproc—dhcp ~f~ DqUeue—dhcp (4.36) IMS Registration Procedure Delay The delay for IMS registration procedure when the registration takes place in 3 G A N is given as: Dimsreg'ig = Ductus—imsreg'Ag ^r Dproc—imsreg ~f" DqUeue—imsreg (4-37) The delay for I M S registration procedure when the registration takes place in W L A N A N is given as: Dimsregwlan D trans—imsregwlan ~~f" Dproc—imsreg ~~f~ DqUeue—imSreg (4.38) IMS Session Setup Delay The delay for I M S session setup when S N is i n U M T S network and C N is in W L A N and vice versa is given by: Dsetup—uw/wu = Dfrans—setup—uw/wu ~r~ Dproc—setup DaUeue—setup (4.39) The delay for IMS session setup when S N as well as C N are in 3 G network is given by: Dsetup—uu = D'trans—setup—uu Dpr0c—setup DqUeue —setup (4.40) 65 Chapter 4. Analysis of SIP-based Signaling for IMS Sessions in 4G Networks The delay for IMS session setup when S N as well as C N are in W L A N is given as: Dsetup—ww — -Dtrans—setup—ww Dproc—setup DqUeue—setup (4-41) Note that if the S N or the C N have not registered wi th the IMS network, then before the IMS session establishment, they have to undergo IMS registration process. In that case, the delay for IMS registration wi l l be added to the total session setup delay. I M S Session Re-establishment Delay To re-establish a session after a S N or C N has undergone a vertical handoff, it needs to send a SIP r e - I N V I T E message to the other terminal. We separately analyze the cases involved in the IMS session re-establishment after a vertical handoff. S N moves to a new 3G network For this case, the steps performed by the S N are: (i) G P R S attach procedure, (ii) P D P context activation procedure, (iii) re-establishment of IMS session using SIP r e - I N V I T E message. W i t h i n this case, further two cases arise, i.e., whether the C N is in W L A N or in U M T S . When the C N is in W L A N , the transmission delay Dvho_uw is given as: Dyho—uw ~ Dgprs—attach ~t~ Dpdp—context "F DSetup—uw (4.42) When the C N is in 3G network, the transmission delay Dvn0_uu is given as: Dvho—uu = Dgprs—attach Dpdp—context ~f" Dsetup—uu (4.43) 66 Chapter 4. Analysis of SIP-based Signaling for IMS Sessions in 4G Networks S N moves to a W L A N For this case, the steps performed by the S N are: (i) D H C P registration procedure, (ii) re-establishment of IMS session using SIP r e - I N V I T E message. W i t h i n this case, further two cases arise, i.e., whether the C N is in W L A N or in U M T S . When the C N is in W L A N , the transmission delay D v f l o _ w w is given as: Dvho—ww = Ddhcp Dsetup—ww (4.44) When the C N is in 3G network, the transmission delay D v h o _ w u is given as: kJyfio—wu Ddfap -\- Dsetup—WU (4.45) 4.2 Numerical Results In this section, we present the numerical results for the delay analysis of SIP-based signaling for IMS sessions. The parameter values selected for the analysis are mentioned hereafter. The value of end-to-end frame propagation delay D for 9.6 kbps, 19.2 kbps, and 128 kbps channel is taken equal to 100 ms whereas for 2 Mbps, and 11 Mbps channel, the valued of D is chosen to be 0.27 ms and 0.049 ms, respectively [23]. Frame duration T as well as inter-frame time r is assumed to be 20 ms for 3G A N [41]. W L A N frame duration is assumed to be 3.5 ms and inter-frame time r is taken to be 1 ms and is independent of the channel bit rate [23]. Total number of users N in the network is taken to be 18600 in accordance wi th the previous chapter. The IP address length L and processor machine word size S are taken to be 32 bits. The system dependent constant value k is equal to 5 [35]. The maximum number of R L P retransmissions n and maximum 67 Chapter 4. Analysis of SIP-based Signaling for IMS Sessions in 4G Networks number of T C P re t ransmiss ions NmaxTcp are b o t h t aken equa l t o 3 [41], [21], [22]. T h e service rate p at a l l the network ent i t ies is t aken equa l to 250 packets/sec. T h e un i t packet process ing cost for a l l the network ent i t ies is t aken equa l to 4 x 1 0 - 3 except the core 3 G network ent i t ies i.e. S G S N a n d G G S N for wh i ch the un i t packet process ing cost is t aken twice as compa red to other network ent i t ies i n accordance w i t h [30], [29]. T h e b a c k g r o u n d u t i l i z a t i o n due to t ra f f i c f r o m other sources is t aken to be 0.7 for H S S a n d A A A server because they have to hand le traf f ic for inter-system commun i ca t i ons f r o m different A N s , 0.5 for the core 3 G ent i t ies i.e. S G S N and G G S N , a n d 0.4 for the rest of the ent i t ies i n accordance w i t h the prev ious chapter . T h e a s sumpt i on of these values of b a c k g r o u n d u t i l i z a t i ons al lows us to determine A e at each of the network nodes. In the first set of exper iments , the I M S reg i s t ra t ion delay, the I M S session setup delay, a n d the I M S session re-establ ishment de lay after S IP-based ver t i ca l handof f are ana l yzed for the channe l rates of 9.6 kbps , 19.2 kbps , a n d 128 kbps i n 3 G network ; and 2 M b p s and 11 M b p s i n W L A N . For th i s set of exper iments , the p robab i l i t y of a f rame be ing i n error p a n d the I M S s igna l ing a r r i va l rate A i n packets per second is kept constant equa l t o 0.02 a n d 9, respect ively. F i gu re 4.6 shows I M S reg i s t ra t ion delay for different channe l rates. It can be seen f r om the f igure tha t for 3 G networks , I M S reg i s t ra t ion delay decreases w i t h the increas ing channe l rates. It can be observed tha t the I M S reg i s t ra t ion delay i n W L A N is cons iderab ly less t h a n the 3 G networks . A l s o , the I M S reg i s t ra t ion delay a lmost rema ins the same i n W L A N independent of the W L A N channe l rate . 68 Chapter 4. Analysis of SIP-based Signaling for IMS Sessions in 4G Networks Figure 4.7 shows the IMS session setup delay when the S N is in 3G network and C N is in W L A N for different combinations of 3G and W L A N channel rates. It can be observed that the IMS session setup delay is greatly effected by the 3 G channel rate i.e. IMS session setup delay decreases considerably as the 3G channel rate increases. Another interesting observation is that the IMS session setup delay is negligibly effected by changing the W L A N channel rate. Figure 4.8 shows the IMS session setup delay when S N as well as C N are in 3 G network for different combination of 3 G channel rates. It can be observed from the figure that the IMS session setup delay decreases with the increasing 3 G channel rates. Figure 4.9 shows the IMS session setup delay when S N as well as C N are in W L A N . It can be observed that the IMS session setup delay in this case is much less as compared to the cases shown in Figure 4.7 and 4.8. Also, there is negligible effect of changing W L A N channel rate in this case on the IMS session setup delay. Figure 4.10 shows the IMS session re-establishment delay when S N moves to a 3 G network and C N is in W L A N for different combination of 3G and W L A N channel rates. It can be observed that the IMS session re-establishment delay decreases with the increasing 3G channel rate and the change in W L A N channel rate has negligible effect on the IMS session re-establishment delay. Figure 4.11 shows the IMS session re-establishment delay when S N moves to a 3G network and C N is in 3G network for different combination of 3G channel rates. It can be seen that the IMS session re-establishment delay decreases when either the S N 3 G 69 Chapter 4. Analysis of SIP-based Signaling for IMS Sessions in 4G Networks channel rate or C N 3 G channel rate or both increases. Figure 4.12 shows the IMS session re-establishment delay when S N moves to a W L A N and C N is in 3G network. It can be observed that the IMS session re-establishment delay decreases wi th the increasing 3 G channel rates. Comparison wi th Figure 4.10 shows that the IMS session re-establishment delay for the case when S N moves to a W L A N and C N is in 3 G network is much less as compared to the case when S N moves to a 3 G network and C N is in W L A N . This is because of different signaling procedures employed in the two cases. Figure 4.13 shows the IMS session re-establishment delay when S N moves to a W L A N and C N is in W L A N . It can be observed that the changing W L A N channel rates has negligible effect in this case. Also comparison with Figure 4.10, 4.11, and 4.12 shows that the IMS session re-establishment delay is minimum for this case. In the second set of experiments, the effect of changing IMS signaling arrival rate A in packets per second on IMS registration delay, IMS session setup delay, and IMS session re-establishment delay after SIP-based vertical handoff in analyzed. The frame error probability p is kept constant at 0.02. The channel considered for 3 G network is 128 kbps and for W L A N is 11 Mbps. Figure 4.14 shows the effect of increasing I M S signaling arrival rate on IMS registration process for 128 kbps 3 G network and 11 Mbps W L A N . It can be seen from the figure that the IMS registration delay increases wi th the increasing arrival rate. Figure 4.15 shows the effect of increasing arrival rate on IMS session setup delay when 70 Chapter 4. Analysis of SIP-based Signaling for IMS Sessions in 4G Networks S N is in 3 G network and C N is in W L A N . It can be seen that the IMS session setup delay increases considerably wi th the increasing arrival rate. Figure 4.16 shows the effect of increasing arrival rate on IMS session re-establishment delay when S N moves to a 3 G network and C N is in W L A N . It can be observed that the IMS session re-establishment delay increases considerably wi th the increasing arrival rates. In the thi rd set of experiments, the effect of changing frame error probability p on IMS registration delay, IMS session setup delay, and IMS session re-establishment delay after SIP-based vertical handoff in analyzed. The arrival rate A is kept constant at 9 packets per second for this set of experiments. The channel considered for 3 G network is 128 kbps and for W L A N is 11 Mbps. Figure 4.17 shows the effect of increasing frame error probability on the IMS registration delay. It can be seen that increasing frame error probability has negligible effect on the IMS registration delay. This is because of the use of R L P which improves the efficiency in the error prone environments as well. Figure 4.18 shows the effect of increasing frame error probability on the IMS session setup delay. It can be observed that the IMS session setup delay increases slowly wi th the increasing frame error probability. The effect of frame error probability is more obvious in IMS session setup delay because of larger number of signaling messages exchanged in the wireless link as compared to the IMS registration process. Figure 4.19 shows the effect of increasing frame error probability on the IMS session re-establishment delay when S N moves to a 128 kbps 3 G network and C N is in 11 Mbps 71 Chapter 4. Analysis of SIP-based Signaling for IMS Sessions in 4G Networks W L A N . It can be noticed that the IMS session re-establishment delay increases slowly with the increasing frame error probability. Comparison wi th Figure 4.18 shows that the IMS session re-establishment delay is effected more by the frame error probability as compared to the IMS session setup delay because of greater number of messages exchanged in the wireless link in the IMS session re-establishment process than the IMS session setup process. 4.3 Summary In this chapter, we analyzed the delay for I M S registration, IMS session establishment and I M S session re-establishment procedures after undergoing a vertical handoff. The delay analysis is comprehensive since it takes into account transmission, processing, and queueing delays at the network entities. Numerical results indicate that increasing the 3G channel rate can significantly decrease the IMS registration, IMS session setup, and IMS session re-establishment delay but changing the W L A N channel rate has negligible effect on the IMS signaling delay. However, the IMS signaling delay in W L A N is much less as compared to that in the 3 G network. The IMS registration, IMS session setup, and IMS session re-establishment delay increase wi th the increase in the arrival rate but are effected negligibly with the increasing frame error probability. 72 Chapter 4. Analysis of SIP-based Signaling for IMS Sessions in 4G Networks UE: User Equipment SGSN: Serving GPRS Support Node GGSN: Gateway GPRS Support Node SGSN: Serving GPRS Support Node DNS: Domain Name System DHCP: Dynamic Host Configuration Protocol PDP: Packet Data Protocol APN: Access Point Name PAP: Password Authentication Protocol CHAP: Challenge Handshake Authentication Protocol DHCP Server 1. Activate PDP Context Request (APN) 2. DNS Response [{GGSN IP Adressj 1 0 . Activate PDP Context Accept «< 2. DNS Query (APN) 4. Create Context Request (PAP, CHAP, PQP PDP j  Request) 9. Create PDP Cbntext Response 5. Radius Authentication Request (PAP, CHAP) 6. Radius Authentication _ Response 7. DHCP Address Request 8. DHCP Address Response Figure 4.2: P D P context activation procedure. UE: User Equipment DHCP: Dynamic Host Configuration Protocol Figure 4.3: D H C P registration process. 73 Chapter 4. Analysis of SIP-based Signaling for IMS Sessions in 4G Networks ( ^ U E ^ ) ( ^ C S C F ^ (^^UCSCF^ ( ^ H S S ^ ) f ^ C S C F ^ ) 1, Register 2. Register 3. Diameter UAR 5. Register 10. 401 Unauthorized 9.401 Unauthorized 4. Diameter UAA 8. 401 6. Diameter MAR 7. Diameter MAA Unauthorized 15. Register ^ 12. Register _ 11. Register ^ 20. 200 OK 19. 200 OK 13. Diameter UAR 14. Diameter UAA 18. 200 OK 16. Diameter SAR 17. Diameter SAA 21. Subscribe 25. Subscribe 22. 200 OK 23. Notify 24. 200 OK 26. Subscribe 28. 200 OK 27. 200 OK 29. Notify 30. Notify 32. 200 OK 31. 200 OK Subscr -iption to reg Event State UE: User Equipment HSS: Home Subscriber Server Diameter UAR: Diameter User-Authentication-Request Diameter UAA: Diameter User-Authentication-Answer Diameter MAR: Diameter Multimedia-Authentication-Request Diameter MAA: Diameter MuStimedia-Authentication-Answer Diameter SAR: Diameter Server-Assignment-Request Diameter SAA: Diameter Server-Assignment-Answer P/I/S-CSCF: Proxy/lnterrogating/Serving-Call Session Control Function F i g u r e 4.4: I M S reg i s t ra t ion process [32], [25]. 74 Chapter 4. Analysis of SIP-based Signaling for IMS Sessions in 4G Networks Originating Originating Visited Home Network Network Interconnection between Operators Terminating Visited Network (5) (S) (S) (S) 1. Invite 2. 100 Trying 20.183 Session Progress 21. PRACK 30. 200 OK 31. Update , 40. 200 OK 46.180 Ringing 47. PRACK 56. 200 OK 62. 200 OK 63. ACK 3. Invite 4.100 Trying 19.183 Session Progress 22. PRACK 29. 200 OK 32. Update^ 39.200 OK 45.180 48. PRACK »• 55. 200 OK 61.200 OK 64. ACK 5. Invite ^ . 100 Trying 18.183 Session Progress 44.180 Ringing 60. 200 OK 7. Diameter LIR •> 8. Diameter LIA 10.100 [Trying 17.183 Session] Progress 23. PRACK 28. 200 OK 33. Update 38. 200 OK 43.180 49. PRACK 54. 200 OK 59.200 65. ACK 9. Invite Ringing OK 11. Invite 12. 100 Trying 16. 183 Session _ Progress 24. PRACK 27. 200 OK 34. Update 37. 200 OK 42,180 Ringing 50. PRACK 53. 200 OK 58. 200 OK 66. ACK 13. Invite 14. 100 Trying 5.183 Sessioil Progress 25. PRACK. 26. 200 OK 35. Update. 36.200 OK ^ 41. 180 „ Ringing 51. PRACK 52. 200 OK 57. 200 OK 67. ACK SN: Source Node CN: Correspondent Node HSS: Home Subscriber Server Diameter LIR: Diameter Location-Information-Request Diameter LIA: Diameter Location-Information-Answer PRACK: Provisional Response Acknowledgment P/I/S-CSCF: Proxy/lnterrogating/Serving-Call Session Control Function F igu re 4.5: I M S session setup procedure [32], [25]. 75 Chapter 4. Analysis of SIP-based Signaling for IMS Sessions in 4G Networks T 3 C o o CD or CO 2 19.2 128 2000 Channel Rates (kbps) 11000 Figure 4.6: IMS registration delay for different channel rates for fixed A and p. 9.6,2 19.2,2 128,2 9.6,11 19.2,11 128,11 SN 3G channel (kbps), CN WLAN channel (Mbps) Figure 4.7: IMS session setup delay for different channel rates when S N is in 3 G network and C N is in W L A N for fixed A and p. 76 Chapter 4. Analysis of SIP-based Signaling for IMS Sessions in 4G Networks Q Q . co 2 co -1 i i - ----: J _ L _. ..__J__t ML 9.6,9.6 19.2,9.6 128,9.6 19.2,19.2 19.2,128 128,128 SN 3G channel (kbps), CN 3G channel (kbps) F igu re 4.8: I M S session setup delay for different channe l rates when S N as wel l as C N are i n 3 G network for f ixed A a n d p. 0.25, 0.05h 2,2 11,2 11,11 SN WLAN channel (Mbps), CN WLAN channel (Mbps) F igu re 4.9: I M S session setup de lay for different channe l rates when S N as wel l as C N are i n W L A N for f ixed A a n d p. 77 Chapter 4. Analysis of SIP-based Signaling for IMS Sessions in 4G Networks I 1—1— a til i I I : 9.6,2 19.2,2 128,2 9.6,11 19.2,11 128,11 SN 3G channel (kbps), CN WLAN channel (Mbps) Figure 4.10: IMS session re-establishment delay for different channel rates when S N moves to a 3G network and C N is in W L A N for fixed A and p. >> 7 0' -iiiiiilif -- -imi -^^^B [flip - -lllitii Wggm 111ft • sfitiiii 9.6,9.6 19.2,9.6 128,9.6 128,19.2 9.6,19.2 9.6,128 19.2,19.219.2,128 128,128 SN 3G channel (kbps), CN 3G channel (kbps) Figure 4.11: IMS session re-establishment delay for different channel rates when S N moves to a W L A N and C N is in 3 G network for fixed A and p. 78 Chapter 4. Analysis of SIP-based Signaling for IMS Sessions in 4G Networks <n >. 2.5 ra a> Q a 1-5 co CD I CD 0> CD CO 0.5 2,9.6 2,19.2 2,128 11,9.6 11,19.2 11,128 SN WLAN channel (Mbps), CN 3G channel (kbps) Figure 4.12: IMS session re-establishment delay for different channel rates when S N moves to a W L A N and C N is in 3 G network for fixed A and p. -s- 0.25 CD e 0.2 0.15 0) o: 0.1 c o in to CD w 0.05 2,2 11,2 2,11 11,11 SN WLAN channel (Mbps), CN WLAN channel (Mbps) Figure 4.13: IMS session re-establishment delay for different channel rates when S N moves to a W L A N and C N is in W L A N for fixed A and p. 79 Chapter 4. Analysis of SIP-based Signaling for IMS Sessions in 4G Networks -*-128 kbps 3G — 11 Mbps WLAN i 1 i i i 0 5 10 15 20 25 IMS Signaling Arrival Rate X (packets per second) Figure 4.14: Effect of changing arrival rate A on IMS registration delay for 128 kbps 3 G network and 11 Mbps W L A N for fixed p. 1.56i . 1 1 1 1 1.54- S -1.52-1.5-1 - 4 8 " 1.46 -1.44J- 1 1 ' 1 0 5 10 15 20 25 IMS Signaling Arrival Rate X (packets per second) Figure 4.15: Effect of changing arrival rate A on IMS session setup delay when S N is in 128 kbps 3 G network and C N is in 11 Mbps W L A N for fixed p. CO c o 0.8 CD Q c o 0.6 g> 0.41 or w 0.2 CD Q CD co CD CO CO 80 Chapter 4. Analysis of SIP-based Signaling for IMS Sessions in 4G Networks 2.98r 2.96-IMS Signaling Arrival Rate X (packets per second) Figure 4.16: Effect of changing arrival rate A on IMS session re-establishment delay when S N moves to 128 kbps 3G network and C N is in 11 Mbps W L A N for fixed p. o CD "5> 0.4 -CD DC ~~ 0.2-OJ ' ' 1 1 1 1 1 1 1 1 0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2 Frame Error Probability p Figure 4.17: Effect of changing frame error probability p on IMS registration delay for 128 kbps 3 G network and 11 Mbps W L A N for fixed A. 81 Chapter 4. Analysis of SIP-based Signaling for IMS Sessions in 4G Networks 1.7, l i i i i i i i i i I 0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2 Frame Error Probability p Figure 4.18: Effect of changing frame error probability p on IMS session setup when S N is in 128 kbps 3G network and C N is in 11 Mbps W L A N for fixed A. 3 .3 r 3 .25-j i 1 1 i i i i i i i I 0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2 Frame Error Probability p Figure 4.19: Effect of changing p on IMS session re-establishment delay when S N moves to 128 kbps 3G network and C N is in 11 Mbps W L A N for fixed A. 82 Chapter 5 Conclusions and Future Work 5.1 Conclusions In this thesis, we proposed two novel interworking architectures for 4 G heterogeneous wireless networks, the L C S C W 2 and T C S C W 2 architectures, based on loosely coupling and tightly coupling paradigm, respectively. The proposed architectures integrate the satellite networks, 3G wireless networks, W i M a x , and W L A N s . The L C S C W 2 architec-ture supports IMS sessions, provides global coverage, and facilitates independent deploy-ment of various access networks. The T C S C W 2 architecture supports IMS sessions and provides global coverage with guaranteed QoS. We also proposed a cost model to deter-mine the associate cost for the IMS signaling and data traffic in the proposed architec-tures. We presented the numerical results for the system cost, as well as the transmission, processing, and queueing costs under different arrival rates and session lengths. We analyzed the delay for IMS registration, IMS session establishment and IMS ses-sion re-establishment procedures after undergoing a vertical handoff in a 4 G environment. The delay analysis takes into account transmission, processing, and queueing delays at the network entities. Numerical results were presented for the delay analysis considering 83 Chapter 5. Conclusions and Future Work different 3G and W L A N channel rates. 5.2 Future Work Based on our proposed architectures, a Hybr id Coupled S a t e l l i t e - C e l l u l a r - W i M a x - W L A N ( H C S C W 2 ) can be proposed that combines the benefits of L C S C W 2 and T C S C W 2 archi-tectures. The H C S C W 2 is envisioned to provide maximum reliability and QoS support, however, the only drawback of H C S C W 2 is the ini t ia l efforts in its deployment. The H C -S C W 2 architecture should be capable of operating in two modes i.e. the tightly coupled mode and the loosely coupled mode. In the H C S C W 2 architecture, there wi l l always be two paths available for traffic, one arising from L C S C W 2 and the other from T C S C W 2 . In the H C S C W 2 architecture, it wi l l be a crit ical decision to choose a path from the two available ones that promises to optimize system performance by providing the least inter-system communication cost. The analytical model proposed for the cost analysis of the interworking architectures can be verified v ia a simulation. The simulation model should confirm the validity of the analytical model. 84 Bibliography [1] 3 G P P . IP Mul t imedia Subsystem (IMS); Stage 2. T S 23.228 (v7.3.0), March 2006. [2] D . Geer. Bui ld ing Converged Networks wi th IMS Technology. IEEE Computer, 38(11):14-16, November 2005. [3] A . K . Salkintzis. W L A N / 3 G Interworking Architecture for Next Generation Hybr id Data Networks. In Proc. of IEEE International Conference on Communications (ICC'04), Paris, France, June 2004. [4] K . D . Wong and V . K . Varma. Supporting Real-Time IP Mul t imedia Services in U M T S . IEEE Communications Magazine, 41(11):148-155, November 2003. [5] S. Tsao and C . L i n . Design and Evaluation of U M T S - W L A N Interworking Strate-gies. In Proc. of IEEE Vehicular Technology Conference (VTC'02), Birmingham, Alabama, September 2002. [6] R. Rajavelsamy, V . Jeedigunta, B . Holur, M . Choudhary, and O. Song. Performance Evaluation of VoIP over 3 G - W L A N Interworking System. In Proc. of IEEE Wireless Communications and Networking Conference (WCNC'05), At lanta , Georgia, March 2005. 85 Bibliography [7] K . Ahmavaara, H . Haverinen, and R. Pichna. Interworking Architecture be-tween 3 G P P and W L A N Systems. IEEE Communications Magazine, 41(11):74-81, November 2003. [8] P. Pinto, L . Bernardo, and P. Sobral. Seamless Continuity of PS-Services in W L A N / 3 G Interworking.. Elsevier Computer Communications, 29(8):1055-1064, M a y 2006. [9] G . Ruggeri, A . Iera, and S. Polito. 802.11-Based Wireless L A N and U M T S Inter-working: Requirements, Proposed Solutions and Open Issues. Elsevier Computer Networks, 47(2):151-166, February 2005. [10] H.S. Mahmood and B . Gage. A n Architecture for Integrating C D M A 2 0 0 0 and 802.11 W L A N Networks. In Proc. of IEEE Vehicular Technology Conference (VTC'03), Orlando, Florida, October 2003. [11] B . G . Evans, M . Werner, E . Lutz , M . Bousquet, G . E . Corazza, G . Mara l , R. Rumeau, and E . Ferro. Integration of Satellite and Terrestrial Systems in Future Mul t imedia Communications. IEEE Wireless Communications, 12(5):72-80, October 2005. [12] D . Ayyagari and A . Ephremides. A Satellite-Augmented Cellular Network Concept. ACM Wireless Networks, 4(2): 189-198, February 1998. [13] V . Y . H . Kueh, R. Tafazolli, and B . G . Evans. Performance Analysis of Session Initi-ation Protocol Based C a l l Set-up over Sa te l l i te -UMTS Network. Elsevier Computer Communications, 28(12):1416-1427, July 2005. 86 Bibliography [14] Z. Sun, D . He, H.S. Cruickshank, L . Liang, A . Sanchez, C. Miguel , V . Schena, C. Tocci , and B . Carro. Scalable Architecture and Evaluation for Mul t ipar ty Con-ferencing over Satellite Links. IEEE Journal on Selected Areas in Communications, 22(3):594-605, A p r i l 2004. [15] L . Fan, R . E . Sheriff, and J . G . Gardiner. Sa te l l i te -UMTS Service Provision Using IP-Based Technology. In Proc. of IEEE Vehicular Technology Conference (VTC'2000), Tokyo, Japan, M a y 2000. [16] V . Y . H . Kueh, N . Wang, and B . Evans. Performance Evaluation of Conference Cre-ation Signalling over Satellite U M T S . In Proc. of IEEE Vehicular Technology Con-ference (VTC'05), Stockholm, Sweden, June 2005. [17] A . Ghosh, D . R . Wolter, J . G . Andrews, and R. Chen. Broadband Wireless Access wi th WiMax/802 .16 : Current Performance Benchmarks and Future Potential. IEEE Communications Magazine, 43(2):129-136, February 2005. [18] Q. Nguyen-Vuong, L . Fiat , and N . Agoulmine. A n Architecture for U M T S - W I M A X Interworking. In Proc. of 1st International Workshop on Broadband Convergence Networks (BcN'06), Vancouver, Canada, A p r i l 2006. [19] D . K i m and A . Ganz. Architecture for 3 G and 802.16 Wireless Networks Integration wi th QoS Support. In Proc. of 2nd International Conference on Quality of Service in Heterogeneous Wired/Wireless Networks (QShine'05), Orlando, Florida, August 2005. 87 Bibliography [20] M . Melrryk and A . Jukan. O n Signaling Efficiency for C a l l Setup in all-IP Wireless Networks. In Proc. of IEEE International Conference on Communications (ICC'06), Istanbul, Turkey, June 2006. [21] H . Fathi , S.S. Chakraborty, and R . Prasad. Optimizat ion of SIP Session Setup Delay for VoIP in 3 G Wireless Networks. IEEE Transactions on Mobile Computing, 5(9): 1121-1132, September 2006. [22] H . Fathi, S.S. Chakraborty, and R. Prasad. O n SIP Session Setup Delay for VoIP Services over Correlated Fading Channels. IEEE Transactions on Vehicular Tech-nology, 55(l):286-295, January 2006. [23] N . Banerjee, W . Wu, K . Basu, and S .K. Das. Analysis of SIP-based Mobi l i ty Manage-ment in 4 G Wireless Networks. Elsevier Computer Communications', 27(8) :697-707, M a y 2004. [24] N . Nakajima, A . Dutta , S. Das, and H . Schulzrinne. Handoff Delay Analysis and Measurement for SIP based mobility in IPv6. In Proc. of IEEE International Con-ference on Communications (ICC03), Anchorage, Alaska, M a y 2003. [25] G . Camari l lo and Miguel-Angel Garcia-Mart in . The 3G IP Multimedia Subsystem (IMS): Merging the Internet and the Cellular Worlds. John Wiley and Sons, 2004. [26] A . K . Salkintzis, C . Fors, and R. Pazhyannur. W L A N - G P R S Integration for Next-Generation Mobile Data Networks. IEEE Wireless Communications, 9(5): 112-124, October 2002. 88 Bibliography [27] M . Buddhikot, G . Chandranmenon, S. Han, Y . W . Lee, S. Mil ler , and L . Salgarelli. Integration of 802.11 and Third-Generation Wireless Da ta Networks. In Proc. of IEEE Conference on Computer Communications (INFOCOM'03), San Francisco, California, March -Apr i l 2003. [28] J . Y . Song, H . J . Lee, S .H. Lee, S.W. Lee, and D . H . Cho. Hybr id Coupling Scheme for U M T S and Wireless L A N Interworking. Elsevier International Journal of Electronics and Communications, July 2006. [29] C . L i u and C . Zhou. A n Improved Interworking Architecture for U M T S - W L A N Tight Coupling. In Proc. of IEEE Wireless Communications and Networking Con-ference (WCNC'05), At lanta , Georgia, March 2005. [30] C . L i u and C . Zhou. H C R A S : A Novel Hybr id Internetworking Architecture between W L A N and U M T S Cellular Networks. In Proc. of IEEE Consumer Communications and Networking Conference (CCNC'05), Las Vegas, Nevada, January 2005. [31] 3 G P P . 3 G P P System to Wireless Loca l Area Network ( W L A N ) Interworking; Sys-tem Description. T S 23.234 (v7.2.0), June 2006. [32] 3 G P P . Signalling flows for the IP multimedia call control based on Session Initiation Protocol (SIP) and Session Description Protocol (SDP); Stage 3 (Release 5). T S 24.228 (v5.15.0), September 2006. [33] P. Calhoun, J . Loughney, E . Gut tman, G . Zorn, and J . Arkko. Diameter Base Protocol. R F C 3588, September 2003. 89 Bibliography [34] J . X i e and L F . Akyi ld iz . A Novel Distributed Dynamic Location Management Scheme for. Min imiz ing Signaling Costs in Mobile IP. IEEE Transactions on Mobile Computing, 1(3):163—175, July-September 2002. [35] B . Lampson, V . Srinivasan, and G . Varghese. IP Lookups Using Mul t iway and Mul t ico lumn Search. IEEE/ACM Transactions on Networking, 7(3):324-334, June 1999. [36] J . Medhi . Stochastic Models in Queueing Theory. Academic Press, A n imprint of Elsevier Science, 2003. [37] E . Gelenbe and G . Pujolle. Introduction to Queueing Networks. John Wiley and Sons, 1998. [38] S. Mohanty and J . X ie . Performance Analysis of a Novel Architecture to Integrate Heterogeneous Wireless Systems. Elsevier Computer Networks, 51 (4): 1095-1105, March 2007. [39] S. Pack and Y . Choi . A Study on Performance of Hierarchical Mobile IPv6 in IP-Based Cellular Networks. IEICE Transactions on Communications, E87-B(3):462-469, March 2004. [40] D . Tarchi, R. Fantacci, and M . Bardazzi . Quali ty of Service Management in I E E E 802.16 Wireless Metropoli tan Area Networks. In Proc. of IEEE International Con-ference on Communications (ICC'06), Istanbul, Turkey, June 2006. 90 Bibliography [41] S .K. Das, E.Lee, K . Basu, and S .K. Sen. Performance Optimizat ion of VoIP Calls over Wireless Links using H.323 Protocol. IEEE Transactions on Comput-ers, 52(6):742-752, June 2003. [42] J .F . Kurose and K . W . Ross. Computer Networking, A Top-Down Approach Featur-ing the Internet. Addison-Wesley, 2003. [43] P. Frojdh, U . Horn, M . Kampmann, A . Nohlgren, and M . Westerlund. Adaptive Streaming within the 3 G P P Packet-Switched Streaming Service. IEEE Network, 20(2):34-40, March -Apr i l 2006. [44] 3 G P P . General Packet Radio Servcie ( G P R S ) ; Service Description; Stage 2. T S 23.060 (v7.3.0), December 2006. [45] R. Droms. Dynamic Host Configuration Protocol. R F C 2131, March 1997. [46] R. Price, C . Bormann, J . Christoffersson, H . Hannu, Z. L i u , and J . Rosenberg. Signaling Compression (SigComp). R F C 3320, January 2003. [47] H . Hannu, J . Christoffersson, S. Forsgren, K . - C . Leung, Z. L i u , and R. Price. Sig-naling Compression (SigComp) - Extended Operations. R F C 3321, January 2003. [48] H . Hannu. Signaling Compression (SigComp) Requirements & Assumptions. R F C 3322, January 2003. [49] G . Camaril lo. Compressing the Session Initiation Protocol (SIP). R F C 3486, Febru-ary 2003. 91 Bibliography [50] H . J in and A C Mahendran. Using SigComp to Compress S I P / S D P Messages. In Proc. of IEEE International Conference on Communications (ICC'05), Seoul, Korea, M a y 2005. [51] J . Rosenberg, H . Schulzrinne, G . Camaril lo, A . Johnston, J . Peterson, R. Sparks, M . Handley, and E . Schooler. SIP: Session Initiation Protocol. R F C 3261, June 2002. 92 

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            data-media="{[{embed.selectedMedia}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
https://iiif.library.ubc.ca/presentation/dsp.831.1-0100992/manifest

Comment

Related Items